PAGERANK BEYOND THE WEB

DAVID F. GLEICH

Abstract. Google’s PageRank method was developed to evaluate the importance of web-pages
via their link structure. The mathematics of PageRank, however, are entirely general and apply to
any graph or network in any domain. Thus, PageRank is now regularly used in bibliometrics, social
and information network analysis, and for link prediction and recommendation. It’s even used for
systems analysis of road networks, as well as biology, chemistry, neuroscience, and physics. We’ll see
the mathematics and ideas that unite these diverse applications.

Key words. PageRank, Markov chain

arXiv:1407.5107v1 [cs.SI] 18 Jul 2014

1. Google’s PageRank. Google created PageRank to address a problem they

encountered with their search engine for the world wide web [Brin and Page, 1998;
Page et al., 1999]. Given a search query from a user, they could immediately find an
immense set of web pages that contained virtually the exact same words as the user
entered. Yet, they wanted to incorporate a measure of a page’s importance into these
results to distinguish highly recognizable and relevant pages from those that were less
well known. To do this, Google designed a system of scores called PageRank that used
the link structure of the web to determine which pages are important. While there are
many derivations of the PageRank equations [Langville and Meyer, 2006; Pan et al.,
2004; Higham, 2005], we will derive it based on a hypothetical random web surfer.
Upon visiting a page on the web, our random surfer tosses a coin. If it comes up heads,
the surfer randomly clicks a link on the current page and transitions to the new page.
If it comes up tails, the surfer teleports to a – possibly random – page independent of
the current page’s identity. Pages where the random surfer is more likely to appear
based on the web’s structure are more important in a PageRank sense.
More generally, we can consider random surfer models on a graph with an arbitrary
set of nodes, instead of pages, and transition probabilities, instead of randomly clicked
links. The teleporting step is designed to model an external influence on the importance
of each node and can be far more nuanced than a simple random choice. Teleporting
is the essential distinguishing feature of the PageRank random walk that had not
appeared in the literature before [Vigna, 2009]. It ensures that the resulting importance
scores always exist and are unique. It also makes the PageRank importance scores
easy to compute.
These features: simplicity, generality, guaranteed existence, uniqueness, and fast
computation are the reasons that PageRank is used in applications far beyond its
origins in Google’s web-search. (Although, the success that Google achieved no
doubt contributed to additional interest in PageRank!) In biology, for instance,
new microarray experiments churn out thousands of genes relevant to a particular
experimental condition. Models such as GeneRank [Morrison et al., 2005] deploy the
exact same motivation as Google, and almost identical mathematics in order to assist
biologists in finding and ordering genes related to a microarray experiment or related
to a disease. Throughout our review, we will see applications of PageRank to biology,
chemistry, ecology, neuroscience, physics, sports, and computer systems.
Two uses underlie the majority of PageRank applications. In the first, PageRank
is used as a network centrality measure [Koschützki et al., 2005]. A network centrality
score yields the importance of each node in light of the entire graph structure. And
the goal is to use PageRank to help understand the graph better by focusing on what
1
2 D. F. Gleich

PageRank reveals as important. It is often compared or contrasted with a host of

other centrality or graph theoretic measures. These applications tend to use global,
near-uniform teleportation behaviors.
In the second type of use, PageRank is used to illuminate a region of a large
graph around a target set of interest; for this reason, we call the second use a localized
measure. It is also called personalized PageRank based on PageRank’s origins in the
web. Consider a random surfer in a large graph that periodically teleports back to
a single start node. If the teleportation is sufficiently frequent, the surfer will never
move far from the start node, but the frequency with which the surfer visits nodes
before teleporting reveals interesting properties of this localized region of the network.
Because of this power, teleportation behaviors are much more varied for these localized
applications.
2. The mathematics of PageRank. There are many slight variations on the
PageRank problem, yet there is a core definition that applies to the almost all of them.
It arises from a generalization of the random surfer idea. Pages where the random
surfer is likely to appear have large values in the stationary distribution of a Markov
chain that, with probability α, randomly transitions according to the link structure of
the web, and with probability 1 − α teleports according to a teleportation distribution
vector v, where v is usually a uniform distribution over all pages. In the generalization,
we replace the notion of “transitioning according to the link structure of the web”
with “transitioning according to a stochastic matrix P.” This simple change divorces
the mathematics of PageRank from the web and forms the basis for the applications
we discuss. Thus, it abstracts the random surfer model from the introduction in a
relatively seamless way. Furthermore, the vector v is a critical modeling tool that
distinguishes between the two typical uses of PageRank. For centrality uses, v will
resemble a uniform distribution over all possibilities; for localized uses, v will focus
the attention of the random surfer on a region of the graph.
Before stating the definition formally, let us fix some notation. Matrices and
vectors are written in bold, Roman letters (A, x), scalars are Greek or indexed, unbold
Roman (α, Ai,j ). The vector e is the column vector of all ones, and all vectors are
column vectors.
Let Pi,j be the probability of transitioning from page j to page i. (Or more
generally, from “thing j” to “thing i”.) The stationary distribution of the PageRank
Markov chain is called the PageRank vector x. It is the solution of the eigenvalue
problem:
(αP + (1 − α)veT )x = x. (2.1)
Many take this eigensystem as the definition of PageRank [Langville and Meyer, 2006].
We prefer the following definition instead:
Definition 2.1 (The PageRank Problem). Let P be a column-stochastic matrix
where all entries are non-negative and the sum of entries in each column is 1. Let
v be a column stochastic vector (eT v = 1), and let 0 < α < 1 be the teleportation
parameter. Then the PageRank problem is to find the solution of the linear system
(I − αP)x = (1 − α)v, (2.2)
where the solution x is called the PageRank vector.
The eigenvector and linear system formulations are equivalent if we seek an
eigenvector x of (2.1) with x ≥ 0 and eT x = 1, in which case:
x = αPx + (1 − α)veT x = αPx + (1 − α)v ⇔ (I − αP)x = (1 − α)v.
 PAGERANK BEYOND THE WEB 3

We prefer the linear system because of the following reasons. In the linear system
setup, the existence and uniqueness of the solution is immediate: the matrix I − αP is
a diagonally dominant M-matrix. The solution x is non-negative for the same reason.
Also, there is only one possible normalization of the solution: x ≥ 0 and eT x = 1.
Anecdotally, we note that, among the strategies to solve PageRank problems, those
based on the linear system setup are both more straightforward and more effective
than those based on the eigensystem approach. And in closing, Page et al. [1999]
describe an iteration more akin to a linear system than an eigenvector.
Computing the PageRank vector x is simple. The humble iteration

x(k+1) = αPx(k) + (1 − α)v where x(0) = v or x(0) = 0

is equivalent both to the power method on (2.1) and the Richardson method on (2.2),
and more importantly, it has excellent convergence properties when α is not too close
to 1. To see this fact, note that the true solution x = αPx + (1 − α)v and consider
the error after a single iteration:

x − x(k+1) = [αPx + (1 − α)v] − [αPx(k) + (1 − α)v] = αP(x − x(k) ).

| {z } | {z }
the true solution x the updated iterate x(k+1)

Thus, the following theorem characterizes the error after k iterations from two different
starting conditions:
Theorem 2.2. Let α, P, v be the data for a PageRank problem to compute
a PageRank vector x. Then the error after k iterations of the update x(k+1) =
αPx(k) + (1 − α)v is:
1. if x(0) = v, then kx − x(k) k1 ≤ kx − vk1 αk ≤ 2αk ; or
2. if x(0) = 0, then the error vector x − x(k) ≥ 0 for all k and kx − x(k) k1 =
eT (x − x(k) ) = αk .
Common values of α range between 0.1 and 0.99; hence, in the worst case,
this method needs at most 3656 iterations to converge to a global 1-norm error of
2−52 ≈ 10−16 (because α3656 ≤ 2−53 to account for the possible factor of 2 if starting
from x(0) = v). For the majority of applications we will see, the matrix P is sparse
with fewer than 10, 000, 000 non-zeros; and thus, these solutions can be computed
efficiently on a modern laptop computer.
Aside 2.3. Although this theorem seems to suggest that x(0) = 0 is a superior
choice, practical experience suggests that starting with x(0) = v results in a faster
method. This may be confirmed by using a computable bound on the error based on
the residual. Let r(k) = (1 − α)v − (I − αP)x(k) = x(k+1) − x(k) be the residual after
k iterations. We can use kx − x(k) k1 = k(I − αP)−1 r(k) k1 ≤ 1−α 1
kr(k) k1 in order to
check for early convergence.
This setup for PageRank, where the choice of P, v, and α vary by application,
applies broadly as the subsequent sections show. However, in many descriptions,
authors are not always careful to describe their contributions in terms of a column
stochastic matrix P and distribution vector v. Rather, they use the following pseudo-
PageRank system instead:
Definition 2.4 (The pseudo-PageRank problem). Let P̄ be a column sub-
stochastic matrix where P̄i,j ≥ 0 and eT P̄ ≤ eT element-wise. Let f be a non-negative
vector, and let 0 < α < 1 be a teleportation parameter. Then the pseudo-PageRank
problem is to find the solution of the linear system

(I − αP̄)y = f (2.3)
4 D. F. Gleich

where the solution y is called the pseudo-PageRank vector.

Again, the pseudo-PageRank vector always exists and is unique because I − αP̄ is
also a diagonally dominant M-matrix. Boldi et al. [2007] was the first to formalize
this definition and distinction between PageRank and pseudo-PageRank, although
they used the term PseudoRank and the normalization (I − αP̄)y = (1 − α)f ; some
advantages of this alternative form are discussed in Section 5.2. The two problems
are equivalent in the following formal sense (which has an intuitive understanding
explained in Section 3.1, Strongly Preferential PageRank):
Theorem 2.5. Let y be the solution of a pseudo-PageRank system with α, P̄
and f . Let v = f /(eT f ). Then if y is renormalized to sum to 1, that is x = y/(eT y),
then x is the solution of a PageRank system with α, P = P̄ + vcT , and v, where
cT = eT − eT P̄ ≥ 0 is a correction vector to make P̄ stochastic.
Proof. First note that α, P, and v is a valid PageRank problem. This is because f
is non-negative and thus v is column stochastic by definition, and also P is column
stochastic because c ≥ 0 (hence P ≥ 0) and eT P = eT P̄ + cT = eT . Next, note that
the solution of the PageRank problem for x satisfies:

αcT x + (1 − α)
x = αP̄x + αvcT x + (1 − α)v = αP̄x + γf where γ= .
eT f
Hence (I − αP̄)x = γf and so x = γy. But, we know that eT x = 1 because x is a
solution of a PageRank problem, and the theorem follows.
The importance of this theorem is it shows that underlying any pseudo-PageRank
system is a true PageRank system in the sense of Definition 2.1. The difference is
entirely in terms of the normalization of the solution – which was demonstrated by Del
Corso et al. [2004]; Berkhin [2005]; Del Corso et al. [2005]. The result of Theorem 2.2
also applies to solving the pseudo-PageRank system, albeit with the following revisions:

Theorem 2.6. Let α, P̄, f be the data for a pseudo-PageRank problem to compute
a pseudo-PageRank vector y. Then the error after k iterations of the update y(k+1) =
αP̄y(k) + f is:
T
1
1. if y(0) = 1−α f , then ky − y(k) k1 ≤ ky − f k1 αk ≤ 2e f k
1−α α ; or
2. if y = 0, then the error vector y − y ≥ 0 for all k and ky − y(k) k1 =
(0) (k)

eT (y − y(k) ) ≤ αk .
Aside 2.7. The error progression proceeds at the same rate for both PageRank
and pseudo-PageRank. This can be improved for pseudo-PageRank if the vector
cT = eT − eT P̄ > 0 (element-wise). In such cases, then we can derive an equivalent
system with a smaller value of α and a suitably rescaled matrix P̄.
These formal results represent the mathematical foundations of all of the PageRank
systems that arise in the literature (with a few technical exceptions that we will study
in Section 5). The results depend only on the construction of a stochastic matrix or
sub-stochastic matrix, a teleportation distribution, and a parameter α. Thus, they
apply generally and have no intrinsic relationship back to the original motivation
of PageRank for the web. Each type of PageRank problem has a unique solution
that always exists, and the two convergence theorems justify that simple algorithms
for PageRank converge to the unique solutions quickly. These are two of the most
attractive features of PageRank.
One final set of mathematical results is important to understand the behavior of
localized PageRank; however, the precise statement of these results requires a lengthy
and complicated diversion into graph partitioning, graph cuts, and spectral graph
 PAGERANK BEYOND THE WEB 5

Fig. 2.1. An illustration of the empirical properties of localized PageRank vectors with teleporta-
tion to a single node in an isolated region. In the graph at left, the teleportation vector is the single
circled node. The PageRank vector is shown as the node color in the right figure. PageRank values
remain high within this region and are nearly zero in the rest of the graph. Theory from Andersen
et al. [2006] explains when this property occurs.

theory. Instead, we’ll state this a bit informally. Suppose that we solve a localized
PageRank problem in a large graph, but the nodes we select for teleportation lie in
a region that is somehow isolated, yet connected to the rest of the graph. Then the
final PageRank vector is large only in this isolated region and has small values on the
remainder of the graph. This behavior is exactly what most uses of localized PageRank
want: they want to find out what is nearby the selected nodes and far from the rest
of the graph. Proving this result involves spectral graph theory, Cheeger inequalities,
and localized random walks – see Andersen et al. [2006] for more detail. Instead, we
illustrate this theory with Figure 2.1.
Next, we will see some of the common constructions of the matrices P and P̄ that
arise when computing PageRank on a graph. These justify that PageRank is also a
simple construction.

3. PageRank constructions. When a PageRank method is used within an

application, there are two common motivations. In the centrality case, the input is
a graph representing relationships or flows between a set of things – they may be
documents, people, genes, proteins, roads, or pieces of software – and the goal is to
determine the expected importance of each piece in light of the full set of relationships
and the teleporting behavior. This motivation was Google’s original goal in crafting
PageRank. In the localized case, the input is also the same type of graph, but the
goal is to determine the importance relative to a small subset of the objects. In
either case, we need to build a stochastic or sub-stochastic matrix from a graph. In
this section, we review some of the common constructions that produce a PageRank
or pseudo-PageRank system. For a visual overview of some of the possibilities, see
Figures 3.1 and 3.2.
Notation for graphs and matrices. Let A be the adjacency matrix for a graph
where we assume that the vertex set is V = {1, . . . , n}. The graph could be directed,
in which case A is non-symmetric, or undirected, in which case A is symmetric. The
graph could also be weighted, in which case Ai,j gives the positive weight of edge (i, j).
Edges with zero weight are assumed to be irrelevant and equivalent to edges that are
not present. For such a graph, let d be the vector of node out-degrees, or equivalently,
the vector of row-sums: d = Ae. The matrix D is simply the diagonal matrix with d
on the diagonal. Weighted graphs are extremely common in applications when the
6 D. F. Gleich

Applications & Theory &

Constructions Algorithms

Strongly preferential Pseudo-

PageRank PageRank
Reverse Dirichlet Eigensystems
Directed or
Random Substochastic Weakly preferential
undirected PageRank
walk matrix PageRank
graph
Linear systems
Weighted Sink preferential
PageRank Other transformations

Fig. 3.1. An overview of PageRank constructions and how they relate. The vast majority of
PageRank applications fall somewhere on the red path.

weights reflect a measure of the strength of the relationships between two nodes.
3.1. The standard random walk. In the standard construction of PageRank,
the matrix P represents a uniform random walk operation on the graph A. When
the graph is weighted, the simple generalization is to model a non-uniform walk that
chooses subsequent nodes with probability proportional to the connecting edge’s weight.
The elements of P̄ are rather similar between the two cases:
Ai,j Ai,j probability of taking the transition
P̄j,i = P = =
k A i,k d i from i to j via a random walk step.

Notice two features of this construction. First, we transpose between j, i and i, j.

This is because Ai,j indicates an edge from node i to node j, whereas the probability
transition matrix element i, j indicates that node i can be reached via node j. Second,
we have written P̄ and P̄j,i here because there may be nodes of the graph with no
outlinks. These nodes are called dangling nodes. Dangling nodes complicate the
construction of stochastic matrices P in a few ways because we must specify a behavior
for the random walk at these nodes in order to fully specify the stochastic matrix.
As a matrix formula, the standard random walk construction is:

P̄ = AT D+ .

Here, we have used the pseudo-inverse of the degree matrix to “invert” the diagonal
matrix in light of the dangling nodes with 0 out-degrees. Let cT be the sub-stochastic
correction vector. For the standard random walk construction, cT is just an indicator
vector for the dangling nodes:
(
X 1 node i is dangling
ci = 1 − P̄k,i =
k
0 otherwise.

We shall now see a few ideas that turn these sub-stochastic matrices into fully
stochastic PageRank problems.
Strongly Preferential PageRank. Given a directed graph with dangling nodes, the
standard random walk construction produces the sub-stochastic matrix P̄ described
above. If we had just used this matrix to solve a pseudo-PageRank problem with
 PAGERANK BEYOND THE WEB 7

a stochastic teleportation vector f = (1 − α)v, then, by Theorem 2.5, the result is

equivalent up to normalization to computing PageRank on the matrix:

P = P̄ + cvT .

This construction models a random walk that transitions according to the distribution
v when visiting a dangling node. This behavior reinforces the effect of the teleportation
vector v, or preference vector as it is sometimes called. Because of this reinforcement,
Boldi et al. [2007] called the construction P = P̄+cvT a strongly preferential PageRank
problem. Again, many authors are not careful to explicitly choose a correction to
turn the sub-stochastic matrix into a stochastic matrix. Their lack of choice, then,
implicitly chooses the strongly preferential PageRank system.
Weakly Preferential PageRank & Sink Preferential PageRank. Boldi et al. [2007]
also proposed the weakly preferential PageRank system. In this case, the behavior
of the random walk at dangling nodes is adjusted independently of the choice of
teleportation vector. For instance, Langville and Meyer [2004] advocates transitioning
uniformly from dangling nodes. In such a case, let u = e/n be the uniform distribution
vector, then a weakly preferential PageRank system is:

P = P̄ + cuT .

We note that another choice of behavior is for the random walk to remain at dangling
nodes until it moves away via a teleportation step:

P = P̄ + diag(c).

We call this final method sink preferential PageRank. These systems are less common.
These choices should be used when the matrix P models some type of information or
material flow that must be decoupled from the teleporting behavior.
3.2. Reverse PageRank. In reverse PageRank, we compute PageRank on the
transposed graph AT . This corresponds to reversing the direction of each edge (i, j) to
be an edge (j, i). Reverse PageRank is often used to determine why a particular node
is important rather than which nodes are important [Fogaras, 2003; Gyöngyi et al.,
2004; Bar-Yossef and Mashiach, 2008]. Intuitively speaking, in reverse PageRank, we
model a random surfer that follows in-links instead of out-links. Thus, large reverse
PageRank values suggest nodes that can reach many nodes in the graph. When these
are localized, they then provide evidence for why a node has large PageRank.
3.3. Dirichlet PageRank. Consider a PageRank problem where we wish to fix
the importance score of a subset of nodes [Chung et al., 2011]. Let S be a subset of
nodes such that i ∈ S implies than vi = 0. A Dirichlet PageRank problem seeks a
solution of PageRank where each node i in S is fixed to a boundary value bi . Formally,
the goal is to find x:

(I − αP)x = (1 − α)v where xi = bi for i ∈ S.

These problems reduce to solving a pseudo-PageRank system. Consider a block

partitioning of P based on the set S and the complement set of vertices S̄:
 
PS,S PS,S̄
P= .
PS̄,S PS̄,S̄
8 D. F. Gleich

Then the Dirichlet PageRank problem is

    
I 0 b 0
= (1 − α) .
−αPS̄,S I − αPS̄,S̄ xS̄ vS̄

This system is equivalent to a pseudo-PageRank problem with P̄ = PS̄,S̄ and f =

(1 − α)vS̄ + αPS̄,S b.
3.4. Weighted PageRank. In the standard random walk construction for Page-
Rank on an unweighted graph, the probability of transitioning from node i to any of
it’s neighbors j is the same: 1/di . Weighted PageRank [Xing and Ghorbani, 2004;
Jiang, 2009] alters this assumption such that the walk preferentially visits high-degree
nodes. Thus, the probability of transitioning from node i to node j depends on the
degree of j relative to the total sum of degrees of all i’s neighbors. In our notation, if
the input is adjacency matrix A with degree matrix D, then the sub-stochastic matrix
P̄ is given by the non-uniform random walk construction on the weighted graph with
adjacency matrix W = AD, that is, P̄ = DAT diag(ADe)+ . More generally, let DW
be a non-negative weighting matrix. It could be derived from the graph itself based
on the out-degree, in-degree, or total-degree (the sum of in- and out-degree), or from
some external source. Then P̄ = DW AT diag(ADW e)−1 . Let us note that weighted
PageRank uses a specific choice of weights for the prior importance of each node; the
setting here already adapts seamlessly to edge-weighted graphs.
3.5. PageRank on an undirected graph. One final construction is to use
PageRank on an undirected graph. Those familiar with Markov chain theory often find
this idea puzzling at first. A uniform random walk on a connected, undirected graph
has a well-known, unique stationary distribution [Stewart, 1994, is a good numerical
treatment of such issues]:
T −1
A D } x = x is solved by x = De/(eT d).
| {z
P

This works because both the row and column sums of A and AT are identical, and the
resulting construction is a reversible Markov chain [Aldous and Fill, 2002, is a good
reference on this topic]. If α < 1, then the PageRank Markov chain is not a reversible
Markov chain even on an undirected graph, and hence, has no simple stationary
distribution. PageRank vectors of undirected graphs, when combined with carefully
constructed teleportation vectors v, yield important information about the presence
of small isolated regions in the graph [Andersen et al., 2006; Gleich and Mahoney,
2014]; formally these results involve graph cuts and small conductance sets. These
vectors are most useful when the teleportation vector is far away from the uniform
distribution, such as the case in Figure 2.1 where the graph is undirected.
Aside 3.1. Of course, if the teleportation distribution v = De/(eT d), then the
resulting chain is reversible. The PageRank vector is then equal to v itself. There
are also specialized PageRank-style constructions that preserve reversibility with more
interesting stationary distributions [Avrachenkov et al., 2010].
4. PageRank applications. When PageRank is used within applications, it
tends to acquire a new name. We will see:
 PAGERANK BEYOND THE WEB 9

3 0

0 0 0 0 0

0
 
1
 
2 1 0 1 0 0 0 2 0
0 0 0 0 1 0 1 0
     
A= d=  c= 
4 5 0 1 1 0 1 0 3 0

0 0 0 0 0 1 1 0
0 0 0 0 1 0 1 0
1 6

A directed graph The adjacency matrix, degree vector, and correction vector

Random walk Strongly preferential Weakly preferential

     
0 1/2 0 0 0 0 0 1/2 0 0 0 0 1/6 1/2 0 0 0 0
 0 0 0 1/3 0 0   0 0 0 1/3 0 0   1/6 0 0 1/3 0 0 
0 1/2 0 1/3 0 0  1/3 1/2 0 1/3 0 0  1/6 1/2 0 1/3 0 0
P̄ =  P= P=
   
 0 0 0 0 0 0 

 1/3 0 0 0 0
 0 

 1/6 0 0 0 0
 0 

0 0 1 1/3 0 1 1/3 0 1 1/3 0 1 1/6 0 1 1/3 0 1
0 0 0 0 1 0 0 0 0 0 1 0 1/6 0 0 0 1 0

P̄ = AT D+ P = P̄ + vcT P = P̄ + ucT
u 6= v

Reverse Dirichlet Weighted

   
0 0 0 0 0 0 0 1/4 0 0 0 0
0 0 1/3 0 0
 
1 0 1/2 0 0 0 0 0 0 3/10 0 0
 1/2 0 1/3 0 0
   
0 0 0 0 1/3 0 0 3/4 0 3/10 0 0
P̄ =  P̄ = P̄ = 
    
0 1 1/2 0 1/3 0   0 0 0 0 0  0 0 0 0 0 0 
  0 1 1/3 0 1  
0 0 0 0 0 1 0 0 1 4/10 0 1
0 0 0 1 0
0 0 0 0 1/3 0 0 0 0 0 1 0
S = {2, 3, 4, 5, 6}
P̄ = A diag(AT e)+ P̄ = P̄S̄,S̄ P̄ = (DW AT ) diag(ADW e)+
S⊂V DW is a diagonal weighting
matrix, e.g. total degree here

Fig. 3.2. A directed graph and some of the different PageRank constructions on that graph.
For the stochastic constructions, we have vT = [ 0 0 13 31 13 0 ] and u = e/n. Note that node 4 is
dangling in the reverse PageRank construction. For the weighted construction, the total degrees are
[ 1 3 3 3 4 2 ].

GeneRank TimedPageRank ObjectRank HostRank

ProteinRank CiteRank FolkRank DirRank
IsoRank AuthorRank ItemRank TrustRank
MonitorRank PopRank BuddyRank BadRank
BookRank FactRank TwitterRank VisualRank

The remainder of this section explores the uses of PageRank within different
domains. It is devoted to the most interesting and diverse uses and should not,
necessarily, be read linearly. Our intention is not to cover the full details, but to survey
the diversity of applications of PageRank. We recommend returning to the primary
sources for additional detail.

Chemistry · §4.1 Literature · §4.7

Biology · §4.2 Bibliometrics · §4.8
Neuroscience · §4.3 Databases & Knowledge systems · §4.9
Engineered systems · §4.4 Recommender systems · §4.10
Mathematical systems · §4.5 Social networks · §4.11
Sports · §4.6 The web, redux · §4.12
10 D. F. Gleich

4.1. PageRank in chemistry. The term “graph” arose from “chemico-graph”

or a picture of a chemical structure [Sylvester, 1878]. Much of this chemical terminology
remains with us today. For instance, the valence of a molecule is the number of potential
bonds it can make. The valence of a vertex is synonymous with its degree, or the
number of connections it makes in the graph. It is fitting, then, that recent work by
Mooney et al. [2012] uses PageRank to study molecules in chemistry. In particular, they
use PageRank to assess the change in a network of molecules linked by hydrogen bonds
among water molecules. Given the output of a molecular dynamics simulation that
provides geometric locations for a solute in water, the graph contains edges between
the water molecules if they have a potential hydrogen bond to a solute molecule. The
goal is to assess the hydrogen bond potential of a solvent. The PageRank centrality
scores using uniform teleportation with α = 0.85 are strongly correlated with the
degree of the node – which is expected – but the deviance of the PageRank score from
the degree identifies important outlier molecules with smaller degree than many in
their local regions. The authors compare the networks based the PageRank values
with and without a solute to find structural differences.
4.2. PageRank in biology & bioinformatics: GeneRank, ProteinRank,
IsoRank. Biology and bioinformatics are currently awash in network data. Some of
the most interesting applications of PageRank arise when it is used to study these
networks. Most of these applications use PageRank to reveal localized information
about the graph based on some form of external data.
GeneRank. Microarray experiments are a measurement of whether or not a gene’s
expression is promoted or repressed in an experimental condition. Microarrays estimate
the outcomes for thousands of genes simultaneously in a few experimental conditions.
The results are extremely noisy. GeneRank [Morrison et al., 2005] is a PageRank-
inspired idea to help to denoise them. The essence of the idea is to use a graph of known
relationships between genes to find genes that are highly related to those promoted or
repressed in the experiment, but were not themselves promoted or repressed. Thus,
they use the microarray expression results as the teleportation distribution vector for
a PageRank problem on a network of known relationships between genes. The network
of relationships between genes is undirected, unweighted with a few thousand nodes.
This problem uses a localized teleportation behaviour and, experimentally, the best
choice of α ranges between 0.75 and 0.85. Teleporting is used to focus the search.
Finding correlated genes. This same idea of using a network of known relationships
in concert with an experiment encapsulates many of the other uses of PageRank in
biology. Jiang et al. [2009] use a combination of PageRank and BlockRank [Kamvar
et al., 2003; Kamvar, 2010] on tissue-specific protein-protein interaction networks in
order to find genes related to type 2 diabetes. The teleportation is provided by 34
proteins known to be related to that disease with α = 0.92.
Winter et al. [2012] use PageRank to study pancreatic ductal adenocarcinoma,
a type of cancer responsible for 130,000 deaths each year, with a particularly poor
prognosis (2% mortality after five years). They identified seven genes that better
predicted patient survival than all existing tools, and validated this in a clinical trial.
One curious feature is that their teleportation parameter was small, α = 0.3. This
was chosen based on a cross-validation strategy in a statistically rigorous way. The
particular type of teleportation they used was based on the correlation between the
expression level of a gene and the survival time of the patient.
ProteinRank. The goal of ProteinRank [Freschi, 2007] is similar, in spirit, to
GeneRank. Given an undirected network of protein-protein interactions and human-
 PAGERANK BEYOND THE WEB 11

curated functional annotations about what these proteins do, the goal is to find proteins
that may share a functional annotation. Thus, the PageRank problem is, again, a
localized use. The teleportation distribution is given by a random choice of nodes with
a specific functional annotation. The PageRank vector reveals proteins that are highly
related to those with this function, but do not themselves have that function labeled.
Protein distance. Recall that the solution of a PageRank problem for a given
teleportation vector v involves solving (I − αP)x = (1 − α)v. The resolvent matrix
X = (1 − α)(I − αP)−1 corresponds to computing PageRank vectors that teleport to
every individual node. The entry Xi,j is the value of the ith node when the PageRank
problem is localized on node j. One interpretation for this score is the PageRank that
node j contributes to node i, which has the flavor of a similarity score between node
i and j. Voevodski et al. [2009] base an affinity measure between proteins on this
idea. Formally, consider an undirected, unweighted protein-protein interaction network.
Compute the matrix X for α = 0.85, and the affinity matrix S = min(X, XT ). (For
an undirected graph, a quick calculation shows that XT = D−1 XD.) For each vertex
i in the graph, form links to the k vertices with the largest values in row of i of S.
These PageRank affinity scores show a much larger correlation with known protein
relationships than do other affinity or similarity metrics between vertices.
IsoRank. Consider the problem of deciding if the vertices of two networks can
be mapped to each other. The relationship between this problem and PageRank is
surprising and unexpected; although precursor literature existed [Jeh and Widom, 2002;
Blondel et al., 2004]. Singh et al. [2007] proposes a PageRank problem to estimate
how much of a match the two nodes are in a diffusion sense. They call it IsoRank
based on the idea of ranking graph isomorphisms. Let P be the Markov chain for
one network and let Q be the Markov chain for the second network. Then IsoRank
solves a PageRank problem on Q ⊗ P. The solution vector x is a vectorized form of a
matrix X where Xij indicates a likelihood that vertex i in the network underlying P
will match to vertex j in the network underlying Q. See Figure 4.1 for an example. If
we have an apriori measure of similarity between the vertices of the two networks, we
can add this as a teleportation distribution term. IsoRank problems are some of the
largest PageRank problems around due to the Kronecker product (e.g. Gleich et al.
[2010] has a problem with 4 billion nodes and 100 billion edges). But there are quite a
few good algorithmic approaches to tackle them by using properties of the Kronecker
product [Bayati et al., 2013] and low-rank matrices [Kollias et al., 2011].
The IsoRank authors consider the problem of matching protein-protein interaction
networks between distinct species. The goal is to leverage insight about the proteins
from a species such as a mouse in concert with a matching between mouse proteins and
human proteins, based on their interactions, in order to hypothesize about possible
functions for proteins in a human. For these problems, each protein is coded by a gene
sequence. The authors construct a teleportation distribution by comparing the gene
sequences of each protein using a tool called BLAST. They found that using α around
0.9 gave the highest structural similarity between the two networks.
4.3. PageRank in neuroscience. The human brain connectome is one of the
most important networks, about which we understand surprisingly little. Applied
network theory is one of a variety of tools currently used to study it [Sporns, 2011].
Thus, it is likely not surprising that PageRank has been used to study the properties of
networks related to the connectome [Zuo et al., 2011]. Most recently, PageRank helped
evaluate the importance of brain regions given observed correlations of brain activity.
In the resulting graph, two voxels of an MRI scan are connected if the correlation
12 D. F. Gleich

0 1/3 1/2 0
" #
1 C 1/2 0 1/2 1
P= A B C D E
B 1/2 1/3 0 0  
 0 1/3 0 0  1 0.03 0.05 0.05 0.09 0.03
0 0 0 1/4 0 2 0.04

0.07 0.07 0.15 0.04

2 3 D 0 0 1/2 1/4 0 3

0.03

0.05 0.05 0.09

0.03
Q=
  
0 1/2 0 1/4 0  4 0.02 0.03 0.03 0.05 0.02
1 1/2 1/2 0 1
4 A E 0 0 0 1/4 0
(1) Two graphs (b) Their stochastic matrices (c) The IsoRank solution

Fig. 4.1. An illustration of the IsoRank problem. The solution, written here as a matrix, gives
the similarity between pairs of nodes of the graph. For instance, node 2 is most similar to node
D. Removing this match, then nodes 1 and 3 are indistinguishable from B and C. Removing these
leaves node 4 equally similar to A and E. In this example we solved (I − αQ ⊗ P)x = (1 − α)e/20
with α = 0.85.

between their functional MRI time-series is high. Edges with weak correlation are
deleted and the remainder are retained with either binary weights or the correlation
weights. The resulting graph is also undirected, and they use PageRank, combined
with community detection and known brain regions, in order to understand changes in
brain structure across a population of 1000 individuals that correlate with age.
Connectome networks are widely hypothesized to be hierarchically organized.
Given a directed network that should express a hierarchical structure, how can we
recover the order of the nodes that minimizes the discrepancy with a hierarchical
hypothesis? Crofts and Higham [2011] consider PageRank for this application on
networks of neural connections from C. Elegans. They find that this gives poor
results compared with other network metrics such as the Katz score [Katz, 1953], and
communicability [Estrada et al., 2008]. In their discussion, the authors note that this
result may have been a mismatch of models, and conjecture that the flow of influence
in PageRank was incorrect. Literature involving Reverse PageRank (Section 3.2)
strengthens this conjecture. Let us reiterate that although PageRank models are easy
to apply, they must be employed with some care in order to get the best results.
4.4. PageRank in complex engineered systems: MonitorRank. The ap-
plications of PageRank to networks in chemistry, biology, and neuroscience are part
of the process of investigating and analyzing something we do not fully understand.
PageRank methods are also used to study systems that we explicitly engineered. As
these engineered systems grow, they become increasingly complex, with networks and
submodules interacting in unpredictable, nonlinear ways. Network analysis methods
like PageRank, then, help reveal these details. We’ll see two examples: software
systems and city systems.
MonitorRank. Diagnosing root causes of issues in a modern distributed system
is painstaking work. It involves repeatedly searching through error logs and tracing
debugging information. MonitorRank [Kim et al., 2013] is a system to provide guidance
to a systems administrator or developer as they perform these activities. It returns a
ranked list of systems based on the likelihood that they contributed to, or participated
in, an anomalous situation. Consider the systems underlying the LinkedIn website:
each service provides one or more APIs that allow other services to utilize its resources.
For instance, the web-page generator uses the database and photo store. The photo
store in turn uses the database, and so on. Each combination of a service and a
programming interface becomes a node in the MonitorRank graph. Edges are directed
and indicate the direction of function calls – e.g. web-page to photo store. Given that
an anomaly was detected in a system, MonitorRank solves a personalized PageRank
 PAGERANK BEYOND THE WEB 13

problem on a weighted, augmented version of the call graph, where the weights and
augmentation depend on the anomaly detected. (The construction is interesting, albeit
tangential, and we refer readers to that paper for the details.) The localized PageRank
scores help determine the anomaly. The graphs involved are fairly small: a few hundred
to a few thousand nodes.
PageRank of the Linux kernel. The Linux kernel is the foundation for an open
source operating system. It has evolved over the past 20 years with contributions from
nearly 2000 individuals in an effort with an estimated value of $3 billion. As of July
2013, the Linux kernel comprised 15.8 million lines of code containing around 300,000
functions. The kernel call graph is a network that represents dependencies between
functions and both PageRank and reverse PageRank, as centrality scores, produce an
ordering of the most important functions in Linux [Chepelianskii, 2010]. The graphs
were directed with a few million edges. Teleportation was typical: α = 0.85 with
a global, uniform v = e/n. They find that utility functions such as printk, which
prints messages from the kernel, and memset, a routine that initializes a region of
memory, have the highest PageRank, whereas routines that initialize the system such
as start kernel have the highest reverse PageRank. Chepelianskii [2010] further
uses the distribution of PageRank and reverse PageRank scores to characterize the
properties of a software system. (This same idea is later used for Wikipedia too, Zhirov
et al. 2010, Section 4.12.)
Roads and Urban Spaces. Another surprising use of PageRank is with road and
urban space networks. PageRank helps to predict both traffic flow and human
movement in these systems. The natural road construction employed is an interesting
graph. A natural road is more or less what it means: it’s a continuous path, built
from road segments by joining adjacent segments together if the angle is sufficiently
small and there isn’t a better alternative. (For help visualizing this idea, consider
traffic directions that state: “Continue straight from High street onto Main street.”
This would mean that there is one natural road joining High street and Main street.)
Using PageRank with α = 0.95, Jiang et al. [2008] finds that PageRank is the best
network measure in terms of predicting traffic on the individual roads. These graphs
have around 15,000 nodes and around 50,000 edges. Another group used PageRank
to study Markov chain models based on the line-graph of roads [Schlote et al., 2012].
That is, given a graph of intersections (nodes) and roads (edges), the line graph, or
dual graph, changes the role of roads to the nodes and intersections to the edges. In
this context, PageRank’s teleportation mirrors the behavior of starting or ending a
journey on each street. This produces a different value of α for each node that reflects
the tendency of individuals to park, or end their journey, on each street. Note that
this is slightly different setup where each node has a separate teleportation parameter
α, rather than a different entry in the teleportation vector. Assuming that each street
has some probability of a journey ending there, then this system is equivalent to a
more general PageRank construction (Section 5.5). These Markov chains are used
to study road planning and optimal routing in light of new constraints imposed by
electric vehicles.
An urban space is the largest space of a city observable from a single vantage
point. For instance, the Mission district of San Francisco is too large, but the area
surrounding Dolores Park is sufficiently small to be appreciated as a whole. For the
study by Jiang [2009], an urban space is best considered as a city neighborhood or block.
The urban space network connects adjacent spaces, or blocks, if they are physically
adjacent. The networks of urban spaces in London, for instance, have up to 20,000
14 D. F. Gleich

nodes and 100,000 links. In these networks, weighted PageRank (Section 3.4) best
predicts human mobility in a case study of movement within London. It outperforms
PageRank, and in fact, they find that weighted PageRank with α = 1 accounts for up
to 60% of the observed movement. Both using weighted PageRank and α = 1 make
sense for these problems – individuals and businesses are likely to co-locate places with
high connectivity, and individuals cannot teleport over the short time-frames used for
the human mobility measurements. Based on the evidence here, we would hypothesize
that using α < 1 would better generalize over longer time-spans.
4.5. PageRank in mathematical systems. Graphs and networks arise in
mathematics to abstract the properties of systems of equations and processes to
relationships between simple sets. We present one example of what PageRank reveals
about a dynamical system by abstracting the phase-space to a discrete set of points and
modeling transitions among them. Curiously, PageRank and its localization properties
has not yet been used to study properties of Cayley graphs from large, finite groups,
although closely related structures have been examined [Frahm et al., 2012].
PageRank of symbolic images and Ulam networks. Let f be a discrete-time
dynamical system on a compact state space M . For instance, M will be the subset of
R2 formed by [0, 2π] × [0, 2π] for our example below. Consider a covering of M by cells
C. In our forthcoming example, this covering will just be a set of non-overlapping cells
that form a regular, discrete partition into cells of size 2π/N × 2π/N . The symbolic
image [Osipenko, 2007] of f with respect to C is a graph where the vertices are the
cells and Ci ∈ C links to Cj ∈ C if x ∈ Ci and f (x) ∈ Cj . The Ulam network is a
weighted approximation to this graph that is constructed by simulating s starting
points within cell Ci and forming weighted links to their destinations Cj [Shepelyansky
and Zhirov, 2010]. The example studied by those authors, and the example we will
consider here, is the Chirikov typical map.
yt+1 = ηyt + k sin(xt + θt )
xt+1 = xt + yt+1 .
It models a kicked oscillator. We generate T random phases θt and look at the map:

f (x, y) = (xT +1 , yT +1 ) mod 2π where x1 = x, y1 = y.

That is, we iterate the map for T steps for each of the T random phase shifts θ1 , . . . , θT .
Applying the construction above with s = 1000 random samples from each cell yields a
directed weighted graph G with N 2 nodes and at most N 2 s edges. PageRank on this
graph, with uniform teleportation, yields beautiful pictures of the transient behaviors
of this chaotic dynamical system; these are easy to highlight with modest teleportation
parameters such as α = 0.85 because this regime inhibits the dynamical system from
converging to its stable attractors. This application is particularly useful for modeling
the effects of different PageRank constructions as we illustrate in Figure 4.2. For that
figure, the graph has 262, 144 nodes and 4, 106, 079 edges, η = 0.99, k = 0.22, T = 10.
4.6. PageRank in sports. Stochastic matrices and eigenvector ranking methods
are nothing new in the realm of sports ranking [Keener, 1993; Callaghan et al., 2007;
Langville and Meyer, 2012]. One of the natural network constructions for sports is
the winner network. Each team is a node in the network, and node i points to node
j if j won in the match between i and j. These networks are often weighted by the
score by which team j beat team i. Govan et al. [2008] used the centrality sense of
PageRank with uniform teleportation and α = 0.85 to rank football teams with these
 PAGERANK BEYOND THE WEB 15

Fig. 4.2. PageRank vectors of the symbolic image, or Ulam network, of the Chirikov typical map
with α = 0.9 and uniform teleportation. From left to right, we show the standard PageRank vector,
the weighted PageRank vector using the unweighted cell in-degree count as the weighting term, and
the reverse PageRank vector. Each node in the graph is a point (x, y), and it links to all other points
(x, y) reachable via the map f (see the text). The graph is weighted by the likelihood of the transition.
PageRank, itself, highlights both the attractors (the bright regions), and the contours of the transient
manifold that leads to the attractor. The weighted vector looks almost identical, but it exhibits an
interesting stippling effect. The reverse PageRank highlights regions of the phase-space that are exited
quickly, and thus, these regions are dark or black in the PageRank vector. The solution vectors were
scaled by the cube-root for visualization purposes. These figures are incredibly beautiful and show
important transient regions of these dynamical systems.

winner networks. The intuitive idea underlying these rankings is that of a random
fan that follows a team until another team beats them, at which point they pick
up the new team, and periodically restarts with an arbitrary team. In the Govan
et al. [2008] construction, they corrected dangling nodes using a strongly preferential
modification, although, we note that a sink preferential modification may have been
more appropriate given the intuitive idea of a random fan. Radicchi [2011] used
PageRank on a network of tennis players with the same construction. Again, this was
a weighted network. PageRank with α = 0.85 and uniform teleportation on the tennis
network placed Jimmy Conors in the best player position.

4.7. PageRank in literature: BookRank. PageRank methods help with three

problems in literature. What are the most important books? Which story paths in
hypertextual literature are most likely? And what should I read next?
For the first question, Jockers [2012] defines a complicated distance metric between
books using topic modeling ideas from latent Dirichlet allocation [Blei et al., 2003].
Using PageRank as a centrality measure on this graph, in concert with other graph
analytic tools, allows Jockers to argue that Jane Austin and Walter Scott are the most
original authors of the 19th century.
Hypertextual literature contains multiple possible story paths for a single novel.
Among American children of similar age to me, the most familiar would be the Choose
your own adventure series. Each of these books consists of a set of storylets; at the
conclusion of a storylet, the story either ends, or presents a set of possibilities for
the next story. Kontopoulou et al. [2012] argue that the random surfer model for
PageRank maps perfectly to how users read these books. Thus, they look for the most
probable storylets in a book. For this problem, the graphs are directed and acyclic,
the stochastic matrix is normalized by outdegree, and we have a standard PageRank
problem. They are careful to model a weakly preferential PageRank system that
deterministically transitions from a terminal (or dangling) storylet back to the start of
16 D. F. Gleich

the book. Teleporting is uniform in their experiments. They find that both PageRank
and a ranking system they derive give useful information about the properties of these
stories.
Books & tags: BookRank. Traditional library catalogs use a carefully curated set
of index terms to indicate the contents of books. These enabled content-based search
prior to the existence of fast full-text search engines. Social cataloging sites such as
LibraryThing and Shelfari allow their users to curate their own set of index terms for
books that they read, and easily share this information among the user sites. The
data on these websites consists of books and tags that indicate the topics of books.
BookRank, which is localized PageRank on the bipartite book-tag graph [Meng, 2009],
produces eerily accurate suggestions for what to read next. For instance, if we use
teleportation to localize on Golub and van Loan’s text “Matrix Computations”, Boyd
and Vandenberghe’s “Convex Optimization”, and Hastie, Tibshirani, and Friedman’s
“Elements of Statistical Learning”, then the top suggestion is a book on Combinatorial
Optimization by Papadimitriou and Steiglitz. A similar idea underlies the general
FolkRank system [Hotho et al., 2006] that we’ll see shortly (Section 4.9).
4.8. PageRank in bibliometrics: TimedPageRank, CiteRank, Author-
Rank. The field of bibliometrics is another big producer and consumer of network
ranking methods, starting with seminal work by Garfield on aggregating data into
a citation network between journals [Garfield, 1955; Garfield and Sher, 1963] and
proceeding through Pinski and Narin [1976], who defined a close analogue of PageRank.
In almost all of these usages, PageRank is used as a centrality measure to reveal the
most important journals, papers, and authors.
Citations among journals. The citation network Garfield originally collected and
analyzed is the journal-journal citation network. It is a weighted network where each
node is a journal and each edge is the number of citations between articles of the
journals. ISI’s impact factor is a more refined analysis of these citation patterns.
Bollen et al. [2006] takes ISI’s methods a step further and finds that a combination of
the impact factor with the PageRank value in the journal citation produces a ranked
list of journals that better correlates with experts’ judgements. PageRank is used as a
centrality measure here with uniform teleportation and weights that correspond to
the weighted citation network. The graph had around 6000 journals. The Eigenfactor
system [West et al., 2010] uses a PageRank vector on the journal co-citation network
with uniform teleportation and α = 0.85 to measure the influence of a journals. It also
shows these rankings on easy-to-browse website.
Citations among papers: TimedPageRank, CiteRank. Moving beyond individual
journals, we can also study the citation network among individual papers using
PageRank. In a paper citation network, each node is an individual article and the
edges are directed based on the citation. Modern bibliographic and citation databases
such as arXiv and DBLP make these networks easy to construct. They tend to have
hundreds of thousands of nodes and a few million edges. TimedPageRank is an idea to
weight the edges of the stochastic matrix in PageRank such that more recent citations
are more important. Formally, it is the solution of

(I − αAT D−1 W)x = (1 − α)e

where W is a diagonal matrix with weights between 0 and 1 that reflects the age of
the paper (1 is recent and 0 is old). The matrix AT D−1 W is column sub-stochastic
and so this is a pseudo-PageRank problem. CiteRank is a subsequent idea that uses
the teleportation in PageRank to increase the rank of recent articles [Walker et al.,
 PAGERANK BEYOND THE WEB 17

2007]. Thus, vi is smaller if paper i is older and vi is larger if paper i is more recent.
The goal of both methods is to produce temporally relevant orderings that remove the
bias of older articles to acquire citations.
While the previous two papers focused on how to make article importance more
accurate, Chen et al. [2007] attempts to use PageRank in concert with the number of
citations to find hidden gems. One notable contribution is the study of α in citation
analysis: based on a heuristic argument about how we build references for an article,
they recommend α = 0.5. Moreover, they find papers with higher PageRank scores
than would be expected given their citation count. These are the hidden gems of the
literature. Ma et al. [2008] uses the same idea in a larger study and find a similar
effect.
Citations among authors: AuthorRank. Another type of bibliographic network is
the co-authorship graph. For each paper, insert edges among all co-authors. Thus,
each paper becomes a clique in the co-authorship network. The weights on each edge
are either uniform (and set to 1), based on the number of papers co-authored, or based
on another weighting construction defined in that paper. All of these constructions
produce an undirected network. PageRank on this network gives a practical ranking
of the most important authors [Liu et al., 2005]. The teleportation is uniform with
α = 0.85, or can be focused on a subset of authors to generate an area-specific ranking.
Their data have a few thousand authors. These graphs are constructions based on an
underlying bipartite matrix B that relates authors and papers. More specifically, the
T
weighted co-authorship network
 0 Bisthe matrix BB . Many such constructions can be
related back to the matrix BT 0 [Dhillon, 2001]. We are not aware of any analysis
that makes a relationship between PageRank in the bipartite graph B0T B 0
and the
weighted matrix BBT .
Author, paper, citation networks. Citation analysis and co-authorship analysis
can, of course, be combined, and that is exactly what Fiala et al. [2008] and Jezek
et al. [2008] do. Whereas Liu et al. [2005] study the co-authorship network, here,
they study a particular construction that joins the bipartite author-paper network to
the citation network to produce an author-citation network. This is a network where
author i links to author j if i has a paper that cites j where j is not a co-author on
that particular paper. Using α = 0.9 and uniform teleportation produces another
helpful list of the most important authors. In the notation of the previous paragraph,
a related construction is the network with adjacency matrix
 
0 B
A= ,
BT C

where B is the bipartite author-paper matrix and C is the citation matrix among
papers. PageRank on these networks takes into account both the co-authorship and
directed citation information, and it rewards authors that have many, highly cited
papers. The graphs studied have a few hundred thousand authors and author-author
citations.
4.9. PageRank in databases and knowledge information systems: Pop-
Rank, FactRank, ObjectRank, FolkRank. Knowledge information systems store
codified forms of information, typically as a relational database. For instance, a
knowledge system about movies consists of relationships between actors, characters,
movies, directors, and so own. Contemporary information systems also often contain
large bodies of user-generated content through tags, ratings, and such. Ratings are a
sufficiently special case that we review them in a forthcoming section (Section 4.10),
18 D. F. Gleich

but we will study PageRank and tags here. PageRank serves important roles as both
a centrality measure and localized measure in networks derived from a knowledge
system. We’ll also present slightly more detail on four interesting applications.
Centrality scores: PopRank, FactRank. PageRank’s role as a centrality measure in
a knowledge information system is akin to its role on the web as an importance measure.
For instance, the authors of PopRank [Nie et al., 2005] consider searching through
large databases of objects – think of academic papers – that have their own internal
set of relationships within the knowledge system – think of co-author relationships.
But these papers are also linked to by websites. PopRank uses web-importance as a
teleportation vector for a PageRank vector defined on the set of object relationships.
The result is a measure of object popularity biased by its web popularity. One of
the challenges in using such a system is that collecting good databases of relational
information is hard. FactRank helps with this process [Jain and Pantel, 2010]. It is a
measure designed to evaluate the importance and accuracy of a fact network. A fact
is just a sentence that connects two objects, such as “David-Gleich wrote the-paper
PageRank-Beyond-The-Web.” These sentences come from textual analysis of large
web crawls. In a fact network, facts are connected if they involve the same set of
objects. Variations on PageRank with uniform teleportation provide lists of important
facts. The authors of FactRank found that weighting relationships between facts and
using PageRank scores of this weighted network gave higher performance than both a
baseline and standard PageRank method in the task of finding correct facts. The fact
networks are undirected and have a few million nodes.
Localized scores: Random-walk with restart, Semi-Supervised Learning. Prediction
tasks akin to the bioinformatics usages of PageRank are standard within knowledge
information systems: networks contain noisy relationships, and the task is inferring,
or predicting, missing data based on these relationships. Zhou et al. [2003] used a
localized PageRank computation to infer the identity of handwritten digits from only a
few examples. These problems were called semi-supervised learning on graphs because
they model the case of finding a vector over vertices (or learning a function) based
on a few values of the function (supervised). It differs from the standard supervised
learning problem because the graph setup implies that only predictions on the vertices
are required, instead of the general prediction problem with arbitrary future inputs. In
the particular study, the graph among these images is based on a radial basis function
construction. For this task α = 0.99 in the pseudo-PageRank system (I − αP)Ỹ = S,
where S is a binary matrix indicating known samples Sij = 1 if image i is known to
be digit j. The largest value in each row of Y = DỸ gives the predicted digit for any
unknown image. While these graphs were undirected, later work [Zhou et al., 2005]
showed how to use PageRank with global teleportation, in concert with symmetric
Laplacian structure defined on a directed graph [Chung, 2005], to enable the same
methodology on a general directed graph.
Pan et al. [2004] define a random walk with restart, which is exactly a personalized
PageRank system, to infer captions for a database of images. Given a set of images
labeled by captions, define a graph where each image is connected to its regions,
each region is connected to other regions via a similarity function, and each image is
connected to the terms in its caption. A query image is a distribution over regions,
and we find terms by solving a PageRank problem with this as the teleportation vector.
These graphs are weighted, undirected graphs. Curiously, the authors chose α based
on experimentation and found that α = 0.1 or α = 0.2 works best. They attribute the
difference to the incredibly small diameter of their network. Subsequent work in the
 PAGERANK BEYOND THE WEB 19

same vein showed some of the relationships with the normalized Laplacian matrix of a
graph [Tong et al., 2006] and returned to a larger value of α around 0.9.
Application 1 – Database queries: ObjectRank. ObjectRank is an interesting type
of database query [Balmin et al., 2004]. A typical query to a database will retrieve all
of the rows of a specified set of tables matching a precise criteria, such as, “find all
students with a GPA of 3.5 that were born in Minnesota.” These tables often have
internal relationships – the database schema – that would help determine which are
the most important returned results. In the ObjectRank model, a user queries the
database with a textual term. The authors describe a means to turn the database
objects and schema into a sub-stochastic transition matrix and define ObjectRank as
the query-dependent solution of the PageRank linear system where the teleportation
vector reflects textual matches. They suggest a great deal of flexibility with defining the
weights of this matrix. For instance, there may be no natural direction for many of these
links and the authors suggest differently weighting forward edges and backward edges
– their intuition is that a paper cited by many important papers is itself important,
but that citing important paper papers does not transfer any importance. They use
α = 0.85 and the graphs have a few million edges.
Application 2 – Folksonomy search: FolkRank. A more specific situation is folk-
sonomy search. A folksonomy is a collection of objects, users, and tags. Each entry is
a triplet of these three items. A user such as myself may have tagged a picture on the
flickr network with the term “sunset” if it contained a sunset, thus creating the triplet
(picture,user,“sunset”). FolkRank scores [Hotho et al., 2006] are designed to measure
the importance of an object, tag, or user with respect to a small set of objects, tags,
or users that define a topic. (This idea is akin to topic-sensitive PageRank, Haveliwala
2002.) These scores then help reveal important objects related to a given search, as
well as the tags that relate them. The scores are based on localized PageRank scores
from an undirected, tripartite weighted network. There is a wrinkle, however. The
FolkRank scores are taken as the difference between a PageRank vector computed
with α = 1 and α = 1/2. The graph is undirected, so the solution with α = 1 is
just the weighted degree distribution. Thus, FolkRank downweights items that are
important for everyone.
Application 3 – Semantic relatedness. The Open Directory Project, or odp, is a
hierarchical, categorical index of web-pages that organizes them into related groups.
Bar-Yossef and Mashiach [2008] suggests a way of defining the relatedness of two
categories on odp using their localized PageRank scores. The goal is to generalize
the idea of the least-common ancestor to random walks to give a different sense of
the distance between categories. To do so, create a graph from the directed hierarchy
in the odp. Let x be the reverse PageRank vector that teleports back to a single
category, and let y be the reverse PageRank vector that teleports back to another
(single) category. Then the relatedness of these categories is the cosine of the angle
between x and y. Let x be the localized PageRank vector (Note the use of reverse
PageRank here so that edges go from child to parent.) They show evidence that this
is a useful measure of relationship in ODP.
Application 4 – Logic programming. A fundamental challenge with scaling logic
programming systems like Prolog is that there is an exponential explosion of potential
combinations and rules to evaluate and, unless the system is extremely well-designed,
these cannot be pruned away. This limits applications to almost trivial problems.
Internally, Prolog-type systems resolve, or prove, logical statements using a search
procedure over an implicitly defined graph that may be infinite. At each node of the
20 D. F. Gleich

graph, the proof system generates all potential neighbors of the node by applying a
rule set given by the logic system. Thus, given one node in the graph, the search
procedure eventually visits all nodes. Localized PageRank provides a natural way
to restrict the search space to only “short” and “likely” proofs [Wang et al., 2013].
Formally, they use PageRank’s random teleportation to control the expansion of the
search procedure. However, there is an intuitive explanation for the random restarts
in such a problem: periodically we all abandon our current line of attack in a proof
and start out fresh. Their system with localized PageRank allows them to realize this
behavior in a rigorous way.
4.10. PageRank in recommender systems: ItemRank. A recommender
system attempts to predict what its users will do based on their past behavior. Netflix
and Amazon have some of the most famous recommendation systems that predict
movies and products, respectively, their users will enjoy. Localized PageRank helps to
score potential predictions in many research studies on recommender systems.
Query reformulation. A key component of modern web-search systems is predicting
future queries. Boldi et al. [2008] run localized PageRank on a query reformulation-
graph that describes how users rewrite queries with α = 0.85. Two queries, q1 and q2 ,
are connected in this graph if a user searched for q1 before q2 within a close time-frame
and both q1 and q2 have some non-trivial textual relationships. This graph is directed
and weighted. The teleportation vector is localized on the current query, or a small set
of previously used terms. PageRank has since had great success for many tasks related
to query suggestion and often performs among the best methods evaluated [Song et al.,
2012].
Item recommendation: ItemRank. Both Netflix and Amazon’s recommender sys-
tems are called item recommendation problems. Users rate items – typically with
a 5-star scale – and we wish to recommend items that a user will rate highly. The
ratings matrix is an items-by-users matrix where Rij is the numeric rating given to
item i by user j. These ratings form a bipartite network between the two groups and
we collapse this to a graph over items as follows. Let G be a weighted graph where
the weights on an edge (i, j) are the number of users that rated both items i and j.
(These weights are equivalent to the number of paths of length 2 between each pair of
items in terms of the bipartite graph.) Let P be the standard weighted random walk
construction on G. Then the ItemRank scores [Gori and Pucci, 2007] are the solutions
of:

(I − αP)S = (1 − α)RD−1
R

where DR are column sums of the rating matrix. Each column of S is a set of
recommendations for user j, and Sij is a proxy for the interest of user j in item i.
Note that any construction of the transition matrix P based on correlations between
items based on user ratings would work in this application as well.
Link prediction. Given the current state of a network, link prediction tries to
predict which edges will come into existence in the future. Liben-Nowell and Kleinberg
[2006] evaluated the localized PageRank score of an unknown edge in terms of its
predictive power. These PageRank values were entries in the matrix (I − αP)−1 for
edges that currently do not exist in the graph. PageRank with α between 0.5 and 0.99
was not one of their best predictors, but the Katz matrix (I−αA)−1 was one of the best
with α = 0.0005. Note that Katz’s matrix is, implicitly, a pseudo-PageRank problem
1
if α < dmax where dmax is the largest degree in the graph. The co-authorship graphs
tested seem to have had degrees less than 2000, making this hidden pseudo-PageRank
 PAGERANK BEYOND THE WEB 21

problem one of the best predictors of future co-authorship. More recent work using
PageRank for predicting links on the Facebook social network includes a training
phase to estimate weights of the matrix P to achieve higher prediction [Backstrom
and Leskovec, 2011]. Localized PageRank is believed to be part of Twitter’s follower
suggestion scheme too [Bahmani et al., 2010].
4.11. PageRank in social networks: BuddyRank, TwitterRank. Page-
Rank serves three purposes in a social network, where the nodes are people and the
edges are some type of social relationship. First, as we discussed in the previous
section, it can help solve link prediction problems to find individuals that will become
friends soon. Second, it serves a classic role in evaluating the centrality of the people
involved to estimate their social status and power. Third, it helps evaluate the potential
influence of a node on the opinions of the network.
Centrality: BuddyRank. Centrality methods have a long history in social networks
– see Katz [1953] and Vigna [2009] for a good discussion. The following claim is difficult
to verify, but we suspect that the first use of PageRank in a large-scale social network
was the BuddyRank measure employed by BuddyZoo in 2003.1 BuddyZoo collected
contact lists from users of the AOL Instant Messenger service and assembled them
into one of the first large-scale social networks studied via graph theoretic methods.
Since then, PageRank has been used to rank individuals in the Twitter network by
their importance [Java, 2007] and to help characterize properties of the Twitter social
network by the PageRank values of their users [Kwak et al., 2010]. These are standard
applications of PageRank with global teleportation and α ≈ 0.85.
Influence. Finding influential individuals is one of the important questions in social
network analysis. This amounts to finding nodes that can spread their influence widely.
More formalizations of this question result in NP-hard optimization problems [Kempe
et al., 2003] and thus, heuristics and approximation algorithms abound [Kempe
et al., 2003, 2005]. Using Reverse PageRank with global teleportation as a heuristic
outperforms out-degree for this task, as shown by Java et al. [2006] for web-blog
influence and Bar-Yossef and Mashiach [2008] for the social network LiveJournal.
Reverse PageRank, instead of traditional PageRank, is the correct model to understand
the origins of influence – the distinction is much like the treatment of hubs and
authorities in other ranking models on networks [Kleinberg, 1999; Blondel et al., 2004].
These ideas also extend to finding topical authorities in social networks by using the
teleportation vector and topic-specific transition probabilities to localize the PageRank
vector in TwitterRank [Weng et al., 2010].
4.12. PageRank in the web, redux: HostRank, DirRank, TrustRank,
BadRank, VisualRank. At the conclusion of our survey of applications, we return
to uses of PageRank on the web itself. Before we begin, let us address the elephant
in the room, so to speak. Does Google still use PageRank? Google reportedly uses a
basket of ranking metrics to determine the final order that results are returned. These
evolve continuously and vary depending on where and when you are searching. It is
unclear to what extent PageRank, or more generally, link analysis measures play a
role in Google’s search ordering, and this is a closely guarded secret unlikely to be
known outside of an inner-circle at Google. One the one hand, in perhaps the only
large-scale published study on PageRank’s effectiveness in a search engine, Najork et al.
[2007] found that it underperformed in-degree. On the other hand, PageRank is still
widely believed to still play some role based on statements from Google. For instance,
1 http://web.archive.org/web/20050724231459/http://buddyzoo.com/
22 D. F. Gleich

Matt Cutts, a Google engineer, wrote about how Google uses PageRank to determine
crawling behavior [Cutts, 2006], and later wrote about how Google moved to a full
substochastic matrix in terms of their PageRank vector [Cutts, 2009]. The latter case
was designed to handle a new class of link on the web called rel=nofollow. This was
an optional HTML parameter that would tell a crawler that the following link is not
useful for relevance judgements. All the major web companies created this parameter
to combat links created in the comment sections of extremely high quality pages such
as the Washington Post. These links are created by users of the Washington Post,
not the staff themselves, and shouldn’t constitute an endorsement on a page. Cutts
described how Google’s new PageRank equation would count these rel=nofollow
links in the degree of a node when it was computing a stochastic normalization, but
would remove the links when computing relevance. For instance, if my page had three
true links and two rel=nofollow links, then my true links would have probabilities
1/5 instead of 1/3, and the sum of my outgoing probability would be 3/5 instead of 1.
Thus, Google’s PageRank computation is a pseudo-PageRank problem now.
Outside of Google’s usage, PageRank is also used to evaluate the web at coarser
levels of granularity through HostRank and DirRank. Reverse PageRank provides
a good measure of a page’s similarity to a hub, according to both Fogaras [2003]
and Bar-Yossef and Mashiach [2008]. PageRank and reverse PageRank also provide
information on the “spaminess” of particular pages through metrics such as TrustRank
and BadRank. PageRank-based information also helped to identify spam directly in a
study by Becchetti et al. [2008]. Finally, PageRank helps identify canonical images to
place on a web-search result (VisualRank).
Coarse PageRank: HostRank, DirRank. Arasu et al. [2002] was an important early
paper that defined HostRank, where the web is aggregated at the level of hostnames.
In this case, all links to and from a hostname, such as www.cs.purdue.edu, become
equivalent. This particular construction models a random surfer that, when visiting a
page, makes a censored, or silent, transition within all pages on the same host, and then
follows a random link. The HostRank scores are the sums of these modified PageRank
scores on the pages within each host [Gleich and Polito, 2007]. Later work included
BlockRank [Kamvar et al., 2003], which used HostRank to initialize PageRank, and
DirRank [Eiron et al., 2004], which forms an aggregation at the level of directories of
websites.
Trust, Reputation, & Spam: TrustRank, BadRank. PageRank typically provides
authority scores to estimate the importance of a page on the web. As the commercial
value of websites grew, it became highly profitable to create spam sites that contain no
new information content but attempt to capture Google search results by appearing
to contain information. BadRank [Sobek, 2003] and TrustRank [Gyöngyi et al., 2004]
emerged as new, link analysis tools to combat the problem. Essentially, these ideas
solve localized, reverse PageRank problems. The results are either used directly, or
as a “safe teleportation” vector for PageRank, as in TrustRank, or in concert with
other techniques, as likely done in BadRank. Kolda and Procopio [2009] generalizes
these models and includes the idea of adding self-links to fix the dangling nodes, like
in sink preferential PageRank, but they add them everywhere, not just at dangling
nodes. For spam-link applications, this way of handling dangling nodes is superior –
in a modeling sense – to the alternatives.
Wikipedia. Wikipedia is often used as a subset of the web for studying ranking. It
is easy to download the data for the entire website, which makes building the web-graph
convenient. (A crawl from a few years ago is in the sparse matrix repository, Davis and
 PAGERANK BEYOND THE WEB 23

Hu 2010, as the matrix Gleich/wikipedia-20070206.) Current graphs of the English

language pages have around 100,000,000 links and 10,000,000 articles. The nature
of the pages on Wikipedia also makes it easy to evaluate results anecdotally. For
instance, we would all raise an eyebrow and demand explanation if “Gene Golub” was
the page with highest global PageRank in Wikipedia. On the other hand, this result
might be expected if we solve a localized PageRank problem around the Wikipedia
article for “numerical linear algebra.” Wissner-Gross [2006] used Wikipedia as a test
set to build reading lists using a combination of localized and global PageRank scores.
Later, Zhirov et al. [2010] computed a 2d ranking on Wikpedia by combining global
PageRank and reverse PageRank. Finally, this 2d ranking showed that Frank Sinatra
was one of the most important people [Eom et al., 2014].
Image search: VisualRank. PageRank also helps to identify “canonical” images to
display as a visual summary of a larger set of images returned from an image search
engine. In the VisualRank system, Jing and Baluja [2008] compute PageRank of
an image similarity graph generated from an image search result. The graphs are
small – around 1000 nodes – which reflects the standard textual query results, and
they are also symmetric and weighted. They solve a global PageRank problem with
uniform teleportation or high-result biased teleportation. The highest ranked images
are canonical images of Mona Lisa amid a diverse collection of views.
5. PageRank generalizations. Beyond the applications discussed so far, there
is an extremely wide set of PageRank-like models that do not fit into the canonical
definition and constructions from Section 3. These support a wide range of additional
applications with mathematics that differs slightly, and some of them are formal
mathematical generalizations of the PageRank vectors. For instance, in prior work, we
studied PageRank with a random teleportation parameter [Constantine and Gleich,
2010]. The standard deviation of these vectors resulted in increased accuracy in
detecting spam pages on the web. We now survey some of these formal generalizations.
5.1. Diffusions, damped sums, & heat kernels. Recall that the pseudo-
PageRank vector is the solution of (2.3),

(I − αP̄)y = f .

Since all of the eigenvalues of P̄ are bounded by 1 in magnitude, the solution y has an
expansion in terms of the Neumann series:
∞
k
X
y= αk P̄ f .
k=0

This expressions gives the pseudo-PageRank vector as a damped sum of powers of P̄

k
where each power, P̄ , has the geometrically decaying weight αk . These are often called
damped diffusions because this equation models how the quantities in f probabilistically
diffuse through the graph where the probability of a path of length k is damped by αk .
Many other sequences serve the same purpose as pointed out by a variety of authors.
Generalized damping. Perhaps the most general setting for these ideas is the
generalized damped PageRank vector:
∞
k
X
z= γk P̄ f (5.1)
k=0
24 D. F. Gleich

P
where γk is a non-negative `1 -sequence (that is, k γk < ∞ and γk ≥ 0). This reduces
to PageRank if γk = αk . Huberman et al. [1998] suggested using such a construction
where γk arises from real-world path following behaviors on the web, which they found
to resemble inverse Gaussian functions. Later results from Baeza-Yates et al. [2006]
proposed essentially the same formula in (5.1). They suggested a variety of interesting
functions γk , including some with only a finite number of non-zero terms. These
authors drew their motivation from the earlier work of TotalRank [Boldi, 2005], which
1 1
suggested γk = k+1 − k+2 in order to evaluate the TotalRank vector:
Z 1
z= (I − αP̄)−1 (1 − α)v dα.
0

This integrates over all possible values of α. (As an aside, this integral is well defined
because a unique limiting PageRank value exists at α = 1, see Section 5.2. This
sidesteps a technical issue with the singular matrix at α = 1.) Our work with making the
value of α in PageRank a random variable is really a further generalization [Constantine
and Gleich, 2010]. Let x(α) be a parameterized form for the PageRank vector for a
fixed graph and teleportation vector. Let A be a random variable supported on [0, 1]
with an infinite number of finite moments, that is, E[Ak ] < ∞ for all k. Intuitively,
A is the probability that a random user of the web follows a link. Our idea was to
use the expected value of PageRank E[x(A)] to produce a ranking that reflected the
distribution of path-following behaviors in the random surfers. We showed:
∞
X
E[x(A)] = (E[Ak ] − E[Ak+1 ])Pk v.
k=0

This results in a family of sequences of γk that depend on the random variable A.

Recent work by Kollias et al. [2013] shows how to evaluate these generalized damped
vectors as a polynomial combination of PageRank vectors in the sense of (2.2).
Heat kernels & matrix exponentials. Another specific case of generalized damping
arises from the matrix exponential, or heat kernel:
∞
X βk k
z = eβ P̄ f = P̄ f .
k!
k=0

Such functions arose in a wide variety of domains that would be tangential to review
here [Estrada, 2000; Miller et al., 2001; Kondor and Lafferty, 2002; Farahat et al.,
2006; Chung, 2007; Kunegis and Lommatzsch, 2009; Estrada and Higham, 2010]. In
terms of a specific relationship with PageRank, Yang et al. [2007] noted that the
pseudo-PageRank vector itself was a single-term approximation to these heat kernel
diffusions. Consider

z = eβ P̄ f ⇔ e−β P̄ z = f ⇔ (I − β P̄ + . . .)z = f .

If we truncate the heat kernel expansion after just the first two terms (I − βP), then we
arise at the pseudo-PageRank vector. (A similar result holds for the formal PageRank
vector too.)
5.2. PageRank limits & eigenvector centrality. In the definition of Page-
Rank used in this paper, we assume that α < 1. PageRank, however, has a unique
well-defined limit as α → 1 [Serra-Capizzano, 2005; Boldi et al., 2005, 2009b]. This
 PAGERANK BEYOND THE WEB 25

is easy to prove using the Jordan canonical form for the case of PageRank (2.2),
but extensions to pseudo-PageRank are slightly more nuanced. As in the previous
section, let x(α) be the PageRank vector as a function of α for a fixed stochastic P:
(I − αP)x(α) = (1 − α)v. Let XJX−1 be the Jordan canonical form of P. Because P
is stochastic, it’s eigenvalues on the unit circle are all semi-simple [Meyer, 2000, page
696]. Thus:
hI i
J= D1 ,
J2

where D1 is a diagonal matrix of the eigenvalues on the unit circle and J2 is a Jordan
block for all eigenvalues with |λ| < 1. We now substitute this into the PageRank
equation:

(I − αP)x(α) = (1 − α)v ⇔ (I − αJ)−1 x̂(α) = (1 − α) |{z}

v̂ .
| {z } −1
X−1 x(α) X v

Using the structure of J decouples these equations:

h I i hI i  x̂(α)0  
v̂0

I −α D1 x̂(α)1 = (1 − α) v̂1 .
I J2 x̂(α)2 v̂2

As α → 1, both x̂(α)1 and x̂(α)2 go to 0 because these linear systems remain non-
singular. Also, note that x̂(α)0 = v̂1 for all α = 6 1, so this point is a removable
singularity. Thus, x̂ can be uniquely defined at α = 1, and hence, so can x. Vigna
[2005] uses the structure of this limit to argue that taking α → 1 in practical applications
is not useful unless the underlying graph is strongly connected, and they propose a new
PageRank construction to ensure this property. Subsequent work by Vigna [2009] does
a nice job of showing how limiting cases of PageRank vectors converge to traditional
eigenvector centrality measures from bibliometrics [Pinski and Narin, 1976] and social
network analysis [Katz, 1953].
The pseudo-PageRank problem does not have nice limiting properties in our
formulation. Let y(α) be a parametric form for the solution of the pseudo-PageRank
system (I − αP̄)y = f . As α → 1, then y → ∞, unless the non-zero support of f
lies outside of a recurrent class, in which case y → 0. Boldi et al. [2005] defines the
PseudoRank system as:

(I − αP̄)y = (1 − α)f

instead. This system always has a non-infinite limit as α → 1. It could, however, have
zero as a limit if P̄ has all eigenvalues less than 1.

5.3. Over-teleportation, negative teleportation, & the Fiedler vector.

The next generalization of PageRank is to values of α > 1. These arose in our prior
work to understand the convergence of quadrature formulas for approximating the
expected value of PageRank with random teleportation parameters [Constantine and
Gleich, 2010]. Mahoney et al. [2012] subsequently showed an amazing relationship
among (i) the Fiedler vector of a graph [Fiedler, 1973; Anderson and Morley, 1985;
Pothen et al., 1990], (ii) a particular generalization of the PageRank vector, which we
call MOV, and (iii) values of α > 1.
26 D. F. Gleich

The Fiedler vector. In contrast to the remainder of this paper, the constructions
and statements in this section are specific to connected, undirected graphs with
symmetric adjacency matrices. The conductance of a set of vertices in a graph is
defined as the number of edges leaving that set, divided by the sum of the degrees of
the vertices within the set. Conductance and its relatives are often used as numeric
quality scores for graph partitioning in parallel processing [Pothen et al., 1990] and
for community detection in graphs [Schaeffer, 2007]. It is NP-hard to find the set of
smallest conductance, but Fiedler’s vector reveals information about the presence of
small conductance sets in a graph through the Cheeger inequality [Chung, 1992]. Let
G be a connected, undirected graph with symmetric adjacency matrix A and diagonal
degree matrix D. The Fiedler vector is the generalized eigenvector of (D−A)q = λ∗ Dq,
with the smallest positive eigenvalue λ∗ > 0. All of the generalized eigenvalues are
non-negative, the smallest is 0, and the largest is bounded above by 1. Cheeger’s
inequality bounds the relationship between λ∗ and the set of smallest conductance in
the graph.
MOV. The MOV vector is defined as the pseudo-inverse solution r in the consistent
linear system of equations:

[(D − A) − γD]r = ρ(γ)Ds, (5.2)

where γ < λ∗ , s is a “seed” vector such that sT De = 0, and ρ(γ) is a scaling constant
such that r has a fixed norm. When γ = 0, this system is singular but consistent,
and thus, we take the pseudo-inverse solution. Note that this is equivalent to the
pseudo-PageRank problem:

(I − αP)z = αρ(γ)f̂
1
where α = 1−γ , z = Dr, and f̂ = Ds. The properties of s in MOV imply that
T
f̂ e = 0, and thus, f̂ must have negative elements, which generalizes the standard
pseudo-PageRank.
In a small surprise, allowing f to take on negative values results in no additional
modeling power in the case of symmetric A. To establish this result, we first observe
that:

(I − αAD−1 ) 1−α
σ
d = σd.

This preliminary fact shows that the pseudo-PageRank vector of an undirected graph
with teleportation according to the degree vector d simply results in a rescaling. We
can use this property to shift any f with negative values in a controlled manner:
σ
(I − αP) (z + 1−α d) = αf̂ + σd,
| {z } | {z }
y f

where σ is chosen such that f ≥ 0 element-wise. Solving these shifted pseudo-PageRank

systems, then, effectively computes the solution z with a well-understood bias term θd.
σ
This is easy to remove afterwards: z = y − 1−α d, at which point we can normalize z
to account for ρ(γ) if desired.
Values of α > 1. While this generalization with negative entries in f gives
no additional mathematical power, it does permit a seamless limit from PageRank
1
vectors to the Fiedler vector. Let α∗ = 1−λ ∗
> 1. The formal result is that the
1
limit limα→α∗ ρ(α) z(α) = q, the Fiedler vector. Note that for the construction of
 PAGERANK BEYOND THE WEB 27

P = AD−1 on an undirected, connected graph, we have that Pk → eT1 d deT as

k → ∞. Thus, when α = 1, the MOV solution z is equivalent to the solution of
(I − (P − eT1 d deT ))z = f because the right hand side of f is orthogonal to the left
eigenvector eT . As all of the eigenvalues of (P − eT1 d deT ) are distinct from 1, this is
a non-singular system. And this fact allows the limit construction to pass through
α = 1 seamlessly. If we additionally assume that f T q 6= 0, then
1
lim z(α) = q,
α→α∗ ρ(α)
and the limiting value of PageRank with over-teleportation is the Fiedler vector. The
analysis in Mahoney et al. [2012], then, interpolates many of the arguments in Vigna
[2009] beyond α = 1 to yield important relationships between spectral graph theory
and PageRank vectors.
5.4. Complex-valued teleportation parameters and a time-dependent
generalization. Again, let x(α) be the PageRank vector (in the sense of (2.2)) as a
function of α for a fixed graph and teleportation vector. Mathematically, the PageRank
vector is a rational function of α. This simple insight produces a host of possibilities,
one of which is evaluating the derivative of the PageRank vector [Boldi et al., 2005;
Golub and Greif, 2006; Gleich et al., 2007]. Another is that PageRank with complex-
valued α is a reasonable mathematical generalization [Horn and Serra-Capizzano, 2007].
Let α ∈ C with |α| < 1, then x(α) has some interesting properties and usages. In
Constantine and Gleich [2010], we needed to bound kx(α)k1 when α was complex. If
α is real and 0 < α < 1, then kx(α)k1 = 1 independent of the choice of α. However, if
|1−α|
α is complex we have: kxk1 ≤ 1−|α| . Later, in Gleich and Rossi [2014], we found that
complex values of α arise in computing closed form solutions to PageRank dynamical
systems where the teleportation vector is a function of time, but the graph remains
fixed. Specifically, the PageRank vector with complex teleportation arises in the
steady-state time-dependent solution of
x0 (t) = (1 − α)v(t) − (I − αP)x(t),
when v(t) oscillates between a fixed set of vectors. Thus, PageRank with complex tele-
portation is both an interesting mathematical problem and has practical applications
in a time-dependent generalization of PageRank.
5.5. Censored node constructions. The final generalized PageRank construc-
tion we wish to discuss is, in fact, a PageRank system hiding inside a Markov chain
construction with a different type of teleportation. In order to motivate the particular
form of this construction, we first review an alternative derivation of the PageRank
vector.
A censored node in a Markov chain is one that exhibits a virtual influence on
the chain in the sense that walks proceed through it as if it were not present. Let
us illustrate this idea by crafting teleportation behavior into a Markov chain in a
different way and computing the PageRank vector itself by censoring that Markov
chain. Suppose that we want to find the stationary distribution of a walk where, if a
surfer wants to teleport, they first transition to a teleport state, and then move from
the teleport state according to the teleportation distribution. The transition matrix of
the Markov chain is:
 
0 αP v
P = .
(1 − α)eT 0
28 D. F. Gleich

And the stationary distribution of this Markov chain is:

v x0
     0
αP x
= , eT x0 + γ = 1.
(1 − α)eT 0 γ γ

Censoring the final teleportation state amounts to modeling its influence on the
stationary distribution, but leaving it with no final contribution. Put more formally,
the stationary distribution of the censored chain is just x0 renormalized to be a
probability distribution: x = x0 /eT x0 . In other words, censoring that state models
pretending that it wasn’t there when determining the stationary distribution, but the
transitions through it still took place; this is equivalent to the standard teleporting
behavior. The vector x is also the PageRank vector of α, P, v, which follows from
1−α 0
x= γ x = 1−α
γ [αPx0 + γv] = αPx + (1 − α)v.

Tomlin [2003], Eiron et al. [2004] and Lee et al. [2007, written in 2003] were some
of the first to observe this property in the context of PageRank; although censoring
Markov chains goes back much further.
There is a more general class of PageRank-style methods that craft transitions
akin to non-uniform teleportation through a censored node construction. Consider,
for example, adding a teleportation node c that connects to all nodes of a network
as in Figure 5.1. This construction gives rise to an implicit PageRank problem with
dmax
α = dmax +1 as we now show. Let
 
A e
A0 =
vT 0

be the adjacency matrix for the modified graph, where v is the teleportation destination
vector. A uniform random walk on this adjacency structure has a single recurrent class,
and thus, a unique stationary distribution [Berman and Plemmons, 1994, Theorem
3.23]. The stationary distribution satisfies:

A (D + I)−1 v/eT v x0
 T     0
0 x
Px=x ⇔ = .
e(D + I)−1 0 γ γ
0
Let P̄ = AT (D + I)−1 . The censored distribution x = x0 /eT x0 is a normalized
solution of the linear system:
0
(I − P̄ )x = v. (5.3)
0
Note that cT = eT − eT P̄ > 0, and so all columns are substochastic. This means that
0
all of the nodes “leak probability” in a semi-formal sense. Scaling P̄ by 1−c1max > 1
adjusts the probabilities such that there is at least one column that is stochastic.
0 0
Consequently, we can write P̄ = αP̄ where α = (1 − cmax ) and P̄ = 1−c1max P̄ . By
substituting this form into (5.3), we have that x is the normalized solution of a
pseudo-PageRank problem where α = 1−c1max . Assuming that A is an unweighted
dmax
graph, then α = dmax +1 .
This idea frequently reappears; for instance, Bini et al. [2010], Lü et al. [2011] and
Schlote et al. [2012] all use it in different contexts.
In a different context, this same type of analysis shows that the Colley matrix
for ranking sports teams is a diagonally perturbed, generalized pseudo-PageRank
 PAGERANK BEYOND THE WEB 29

3
2 
0 1/3 0 0 0 0
 
1

0 0 0 1/4 0 0  1/3
4 5 0 1/3 0 1/4 0 0  1/2
   
P̄ =  c=
0 0 0 0 0 0  1/4
 
0 0 1/2 1/4 0 1/2 1/2
1 6 0 0 0 0 1/2 0 1/2
c

(a) A directed graph with a (b) The substochastic matrix and correction vector for the Markov
censored node c chain construction after node c is censored.

Fig. 5.1. In this teleportation construction we add a node c to the original graph as in subfigure
(a). The probability of transitioning to c, or teleporting after we censor node c, then depends on the
degree of each node. A random surfer teleports from node 2 with probability 1/3 and from node 4
with probability 1/4. This construction yields a substochastic matrix P̄ where all the elements of
the correction vector c are positive. This means it’s equivalent to a PageRank construction with
α = 1 − min c, or α = 3/4 for this problem.

system [Colley, 2002; Langville and Meyer, 2012]. Let the symmetric, weighted graph
G represent the network of times team i played team j. And let f be a vector of the
accumulated scores differences over all of those games. It could have negative entries,
rendering it outside of our traditional framework, however, as we saw in Section 5.3,
this is a technical detail that is avoidable. The vector of Colley scores r is the solution
of:
(D + 2I − A)r = f .
Let y = (D + 2I)−1 r. Then,
(I − αP̄)y = f
dmax
where α = dmax +2 . This analysis establishes a formal relationship between Markov
style ranking metrics [Langville and Meyer, 2012] and the least-squares style ranking
metrics employed by Colley. It also enables us to use fast PageRank solvers for these
Colley systems.
6. Discussion & a positive outlook on PageRank’s wide usage. PageRank
has gone from being used to evaluate the importance of web pages to a much broader
set of applications. The method is easy to understand, is robust and reliable, and
converges quickly. Most applications solve PageRank problems of only a modest size,
with fewer than 100,000,000 vertices; this regime permits a much wider variety of
algorithmic possibilities than those that must only work on the web.
We have avoided the discussion of PageRank algorithms entirely in this manuscript
because, by and large, simple iterations suffice for fast convergence in this regime.
Values of α tend to be less than 0.99, which requires fewer than 2000 iterations to
converge to machine precision. Nevertheless, there is ample opportunity to accelerate
PageRank computations in this regime as there are ideas that involve computing
multiple PageRank vectors for a single task. One example is PerturbationRank [Du
et al., 2008], which uses the perturbation induced in a PageRank vector by removing a
node to compute a new ranking of all nodes. Thus, innovations in PageRank algorithms
are still relevant, but must be made within the context of these small-scale uses.
There are also a great number of PageRank-like ideas outside of our specific canon.
For instance, none of the following models fit our PageRank framework:
30 D. F. Gleich

BrowseRank Liu et al. [2008] define a continuous time Markov chain to model a
random surfer that remains on a specified node for some period of time before
transitioning away. This model handles sites like Facebook, where users spend
significant time interacting within a single page.
Voting Boldi et al. [2009a] and Boldi et al. [2011] define a voting model on a social
network inspired by computing Katz or PageRank on a random network where
each node picks a single outlink.
SimRank This problem is another way to use PageRank-like ideas to evaluate sim-
ilarity between the nodes of a graph (like the IsoRank problem) [Jeh and
Widom, 2002]. SimRank, however, involves solving a linear system on a row
sub-stochastic matrix.
Food webs The food web is a network where species are linked by the feeding
relationships. Allesina and Pascual [2009] point out a few modifications to
PageRank to make it more appropriate. First, they use teleportation to
model a constant loss of nutrients from higher-level species and reinject these
nutrients through primary producers (such as bacteria). Second, they note
that the flow of importance ought to be reversed so that species i points
to species j if i is important for j’s survival. The result is an eigenvector
computation on a fully stochastic matrix.
Opinion dynamics Models of opinion formation on social network posit strikingly
similar dynamics to a PageRank iteration [Friedkin and Johnsen, 1990, 1999].
The essential difference is that a node’s opinion is the average of its in-links,
instead of propagating its value to its out-links. Like SimRank, this results in
a row sub-stochastic iteration.
The details and implications of these models are fascinating, and this manuscript
would double in size if we were to treat them.
In most successful studies, PageRank is used as a type of baseline measure. Its
widespread success above extremely simple baselines suggests that its modified random
walk is a generally useful alternative worth investigating. In this sense, it resembles a
form of regularization. And this is how we feel that PageRank should be used. Note
that studies must use care when determining the type of PageRank construction –
weighted, reverse, Dirichlet, etc. – as this can make a large difference in the quality
of the results. Consider, for instance, the use of weighted PageRank in Jiang [2009].
In their application, they wanted to model where people move, and it makes good
sense that businesses would locate in places with many connections and therefore, that
people would preferentially move to these same locations. Given the generality of the
idea and its intuitive appeal, we anticipate continued widespread use of the PageRank
idea over the next 20 years in new and exciting applications as network data continues
to proliferate.

REFERENCES

D. Aldous and J. A. Fill. Reversible Markov chains and random walks on graphs. 2002. Unfinished
monograph, recompiled 2014, available at http://www.stat.berkeley.edu/$\sim$aldous/RWG/
book.html. Cited on page 8.
S. Allesina and M. Pascual. Googling food webs: Can an eigenvector measure species’ importance
for coextinctions? PLoS Comput Biol, 5 (9), p. e1000494, 2009. doi:10.1371/journal.pcbi.
1000494. Cited on page 30.
R. Andersen, F. Chung, and K. Lang. Local graph partitioning using PageRank vectors. In
Proceedings of the 47th Annual IEEE Symposium on Foundations of Computer Science. 2006.
Cited on pages 5 and 8.
 PAGERANK BEYOND THE WEB 31

W. N. J. Anderson and T. D. Morley. Eigenvalues of the Laplacian of a graph. Linear and

Multilinear Algebra, 18 (2), pp. 141–145, 1985. doi:10.1080/03081088508817681. Cited on
page 25.
A. Arasu, J. Novak, A. Tomkins, and J. Tomlin. PageRank computation and the structure of the
web: Experiments and algorithms. In Proceedings of the 11th international conference on the
World Wide Web. 2002. Poster session. Cited on page 22.
K. Avrachenkov, B. Ribeiro, and D. Towsley. Improving random walk estimation accuracy with
uniform restarts. In Algorithms and Models for the Web-Graph, pp. 98–109. Springer Berlin
Heidelberg, 2010. doi:10.1007/978-3-642-18009-5_10. Cited on page 8.
L. Backstrom and J. Leskovec. Supervised random walks: Predicting and recommending links in
social networks. In Proceedings of the Fourth ACM International Conference on Web Search
and Data Mining, pp. 635–644. 2011. doi:10.1145/1935826.1935914. Cited on page 21.
R. Baeza-Yates, P. Boldi, and C. Castillo. Generalizing PageRank: Damping functions for
link-based ranking algorithms. In Proceedings of the 29th annual international ACM SIGIR
conference on Research and development in information retrieval (SIGIR2006), pp. 308–315.
2006. doi:10.1145/1148170.1148225. Cited on page 24.
B. Bahmani, A. Chowdhury, and A. Goel. Fast incremental and personalized pagerank . Proc.
VLDB Endow., 4 (3), pp. 173–184, 2010. Cited on page 21.
A. Balmin, V. Hristidis, and Y. Papakonstantinou. ObjectRank: authority-based keyword search
in databases. In Proceedings of the Thirtieth international conference on Very large data bases
- Volume 30, pp. 564–575. 2004. Cited on page 19.
Z. Bar-Yossef and L.-T. Mashiach. Local approximation of PageRank and reverse PageRank . In
CIKM ’08: Proceeding of the 17th ACM conference on Information and knowledge management,
pp. 279–288. 2008. doi:10.1145/1458082.1458122. Cited on pages 7, 19, 21, and 22.
M. Bayati, D. F. Gleich, A. Saberi, and Y. Wang. Message-passing algorithms for sparse network
alignment. ACM Trans. Knowl. Discov. Data, 7 (1), pp. 3:1–3:31, 2013. doi:10.1145/2435209.
2435212. Cited on page 11.
L. Becchetti, C. Castillo, D. Donato, R. Baeza-Yates, and S. Leonardi. Link analysis for
web spam detection. ACM Trans. Web, 2 (1), pp. 1–42, 2008. doi:10.1145/1326561.1326563.
Cited on page 22.
P. Berkhin. A survey on PageRank computing. Internet Mathematics, 2 (1), pp. 73–120, 2005.
Cited on page 4.
A. Berman and R. J. Plemmons. Nonnegative Matrices in the Mathematical Sciences, SIAM, 1994.
Cited on page 28.
D. A. Bini, G. M. D. Corso, and F. Romani. A combined approach for evaluating papers,
authors and scientific journals. Journal of Computational and Applied Mathematics, 234 (11),
pp. 3104 – 3121, 2010. Numerical Linear Algebra, Internet and Large Scale Applications.
doi:10.1016/j.cam.2010.02.003. Cited on page 28.
D. M. Blei, A. Y. Ng, and M. I. Jordan. Latent Dirichlet allocation. Journal of Machine Learning
Research, 3, pp. 993–1022, 2003. Cited on page 15.
V. D. Blondel, A. Gajardo, M. Heymans, P. Senellart, and P. V. Dooren. A measure of
similarity between graph vertices: Applications to synonym extraction and web searching. SIAM
Review, 46 (4), pp. 647–666, 2004. doi:10.1137/S0036144502415960. Cited on pages 11 and 21.
P. Boldi. TotalRank: Ranking without damping. In Poster Proceedings of the 14th international
conference on the World Wide Web (WWW2005), pp. 898–899. 2005. Cited on page 24.
P. Boldi, F. Bonchi, C. Castillo, D. Donato, A. Gionis, and S. Vigna. The query-flow graph:
Model and applications. In Proceedings of the 17th ACM Conference on Information and
Knowledge Management, pp. 609–618. 2008. doi:10.1145/1458082.1458163. Cited on page 20.
P. Boldi, F. Bonchi, C. Castillo, and S. Vigna. Voting in social networks. In CIKM ’09:
Proceeding of the 18th ACM conference on Information and knowledge management, pp.
777–786. 2009a. doi:10.1145/1645953.1646052. Cited on page 30.
———. Viscous democracy for social networks. Commun. ACM, 54 (6), pp. 129–137, 2011.
doi:10.1145/1953122.1953154. Cited on page 30.
P. Boldi, R. Posenato, M. Santini, and S. Vigna. Traps and pitfalls of topic-biased PageRank .
In WAW2006, Fourth International Workshop on Algorithms and Models for the Web-Graph,
pp. 107–116. 2007. doi:10.1007/978-3-540-78808-9_10. Cited on pages 4 and 7.
P. Boldi, M. Santini, and S. Vigna. PageRank as a function of the damping factor . In Proceedings
of the 14th international conference on the World Wide Web (WWW2005). 2005. doi:
10.1145/1060745.1060827. Cited on pages 24, 25, and 27.
———. PageRank: Functional dependencies. ACM Trans. Inf. Syst., 27 (4), pp. 1–23, 2009b.
doi:10.1145/1629096.1629097. Cited on page 24.
J. Bollen, M. A. Rodriquez, and H. Van de Sompel. Journal status. Scientometrics, 69 (3), pp.
32 D. F. Gleich

669–687, 2006. doi:10.1007/s11192-006-0176-z. Cited on page 16.

S. Brin and L. Page. The anatomy of a large-scale hypertextual web search engine. Comput. Netw.
ISDN Syst., 30 (1-7), pp. 107–117, 1998. doi:10.1016/S0169-7552(98)00110-X. Cited on page
1.
T. Callaghan, P. J. Mucha, and M. A. Porter. Random walker ranking for NCAA division I-A
football. The American Mathematical Monthly, 114 (9), pp. 761–777, 2007. Cited on page 14.
P. Chen, H. Xie, S. Maslov, and S. Redner. Finding scientific gems with Google’s PageRank
algorithm. Journal of Informetrics, 1 (1), pp. 8–15, 2007. doi:10.1016/j.joi.2006.06.001.
Cited on page 17.
A. Chepelianskii. Towards physical laws for software architecture. arXiv, cs.SE, p. 1003.5455,
2010. Cited on page 13.
F. Chung. Laplacians and the Cheeger inequality for directed graphs. Annals of Combinatorics,
9 (1), pp. 1–19, 2005. 10.1007/s00026-005-0237-z. doi:10.1007/s00026-005-0237-z. Cited on
page 18.
———. The heat kernel as the PageRank of a graph. Proceedings of the National Academy of
Sciences, 104 (50), pp. 19735–19740, 2007. doi:10.1073/pnas.0708838104. Cited on page 24.
F. Chung, A. Tsiatas, and W. Xu. Dirichlet pagerank and trust-based ranking algorithms. In
Algorithms and Models for the Web Graph, pp. 103–114. Springer Berlin Heidelberg, 2011.
doi:10.1007/978-3-642-21286-4_9. Cited on page 7.
F. R. L. Chung. Spectral Graph Theory, American Mathematical Society, 1992. Cited on page 26.
W. N. Colley. Colley’s bias free college football ranking method: The Colley matrix explained.
Technical report, Princeton University, 2002. Cited on page 29.
P. G. Constantine and D. F. Gleich. Random alpha PageRank . Internet Mathematics, 6 (2), pp.
189–236, 2010. doi:10.1080/15427951.2009.10129185. Cited on pages 23, 24, 25, and 27.
J. J. Crofts and D. J. Higham. Googling the brain: Discovering hierarchical and asymmetric
network structures, with applications in neuroscience. Internet Mathematics, 7, pp. 233–254,
2011. doi:10.1080/15427951.2011.604284. Cited on page 12.
M. Cutts. Q&A march 27, 2006 . Matt Cutts: Gadgets, Google, and SEO. Available online
http://www.mattcutts.com/blog/q-a-thread-march-27-2006/, 2006. Cited on page 22.
———. PageRank scupting. Matt Cutts: Gadgets, Google, and SEO blog. Available online
http://www.mattcutts.com/blog/pagerank-sculpting/, 2009. Cited on page 22.
T. A. Davis and Y. Hu. The University of Florida sparse matrix collection. ACM Transactions on
Mathematical Software, 2010. To appear. Cited on page 22.
G. Del Corso, A. Gullı́, and F. Romani. Fast PageRank computation via a sparse linear system
(extended abstract). In Algorithms and Models for the Web-Graph, pp. 118–130. Springer Berlin
Heidelberg, 2004. doi:10.1007/978-3-540-30216-2_10. Cited on page 4.
G. M. Del Corso, A. Gullı́, and F. Romani. Fast PageRank computation via a sparse linear
system. Internet Mathematics, 2 (3), pp. 251–273, 2005. Cited on page 4.
I. S. Dhillon. Co-clustering documents and words using bipartite spectral graph partitioning. In
Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery
and data mining, pp. 269–274. 2001. doi:10.1145/502512.502550. Cited on page 17.
Y. Du, J. Leung, and Y. Shi. PerturbationRank: A non-monotone ranking algorithm. Technical
report, University of Michigan, 2008. Cited on page 29.
N. Eiron, K. S. McCurley, and J. A. Tomlin. Ranking the web frontier . In Proceedings of
the 13th international conference on the World Wide Web (WWW2004), pp. 309–318. 2004.
doi:10.1145/988672.988714. Cited on pages 22 and 28.
Y.-H. Eom, P. Aragón, D. Laniado, A. Kaltenbrunner, S. Vigna, and D. L. Shepelyansky.
Interactions of cultures and top people of Wikipedia from ranking of 24 language editions. arXiv,
cs.SI, p. 1405.7183, 2014. Cited on page 23.
E. Estrada. Characterization of 3d molecular structure. Chemical Physics Letters, 319 (5-6), pp.
713–718, 2000. doi:10.1016/S0009-2614(00)00158-5. Cited on page 24.
E. Estrada and D. J. Higham. Network properties revealed through matrix functions. SIAM
Review, 52 (4), pp. 696–714, 2010. doi:10.1137/090761070. Cited on page 24.
E. Estrada, D. J. Higham, and N. Hatano. Communicability and multipartite structures in
complex networks at negative absolute temperatures. Phys. Rev. E, 78, p. 026102, 2008.
doi:10.1103/PhysRevE.78.026102. Cited on page 12.
A. Farahat, T. LoFaro, J. C. Miller, G. Rae, and L. A. Ward. Authority rankings from HITS,
PageRank, and SALSA: Existence, uniqueness, and effect of initialization. SIAM Journal on
Scientific Computing, 27 (4), pp. 1181–1201, 2006. doi:10.1137/S1064827502412875. Cited on
page 24.
D. Fiala, F. Rousselot, and K. Jeek. PageRank for bibliographic networks. Scientometrics, 76 (1),
pp. 135–158, 2008. doi:10.1007/s11192-007-1908-4. Cited on page 17.
 PAGERANK BEYOND THE WEB 33

M. Fiedler. Algebraic connectivity of graphs. Czechoslovak Mathematical Journal, 23 (98), pp.

298–305, 1973. Cited on page 25.
D. Fogaras. Where to start browsing the web? In Innovative Internet Community Systems, pp.
65–79. Springer Berlin Heidelberg, 2003. doi:10.1007/978-3-540-39884-4_6. Cited on pages
7 and 22.
K. M. Frahm, A. D. Chepelianskii, and D. L. Shepelyansky. PageRank of integers. Journal of
Physics A: Mathematical and Theoretical, 45 (40), p. 405101, 2012. doi:10.1088/1751-8113/
45/40/405101. Cited on page 14.
V. Freschi. Protein function prediction from interaction networks using a random walk ranking
algorithm. In Proceedings of the 7th IEEE International Conference on Bioinformatics and
Bioengineering (BIBE 2007), pp. 42–48. 2007. doi:10.1109/BIBE.2007.4375543. Cited on page
10.
N. E. Friedkin and E. C. Johnsen. Social influence and opinions. The Journal of Mathematical
Sociology, 15 (3-4), pp. 193–206, 1990. doi:10.1080/0022250X.1990.9990069. Cited on page
30.
———. Social influence networks and opinion change. Advances in Group Processes, 16 (1), pp.
1–29, 1999. Cited on page 30.
E. Garfield. Citation indexes for science: A new dimension in documentation through association
of ideas. Science, 122 (3159), pp. 108–111, 1955. doi:10.1126/science.122.3159.108. Cited
on page 16.
E. Garfield and I. H. Sher. New factors in the evaluation of scientific literature through citation
indexing. American Documentation, 14 (3), pp. 195–201, 1963. doi:10.1002/asi.5090140304.
Cited on page 16.
D. F. Gleich, P. Glynn, G. H. Golub, and C. Greif. Three results on the PageRank vector:
eigenstructure, sensitivity, and the derivative. In Web Information Retrieval and Linear Algebra
Algorithms. 2007. Cited on page 27.
D. F. Gleich, A. P. Gray, C. Greif, and T. Lau. An inner-outer iteration for PageRank . SIAM
Journal of Scientific Computing, 32 (1), pp. 349–371, 2010. doi:10.1137/080727397. Cited on
page 11.
D. F. Gleich and M. M. Mahoney. Algorithmic anti-differentiation: A case study with min-cuts,
spectral, and flow. In Proceedings of the International Conference on Machine Learning (ICML).
2014. Cited on page 8.
D. F. Gleich and M. Polito. Approximating personalized PageRank with minimal use of webgraph
data. Internet Mathematics, 3 (3), pp. 257–294, 2007. doi:10.1080/15427951.2006.10129128.
Cited on page 22.
D. F. Gleich and R. A. Rossi. A dynamical system for PageRank with time-dependent teleportation.
Internet Mathematics, 10 (1–2), pp. 188–217, 2014. doi:10.1080/15427951.2013.814092. Cited
on page 27.
G. Golub and C. Greif. An Arnoldi-type algorithm for computing PageRank . BIT Numerical
Mathematics, 46 (4), pp. 759–771, 2006. doi:10.1007/s10543-006-0091-y. Cited on page 27.
M. Gori and A. Pucci. ItemRank: a random-walk based scoring algorithm for recommender engines.
In IJCAI’07: Proceedings of the 20th international joint conference on Artifical intelligence,
pp. 2766–2771. 2007. Cited on page 20.
A. Y. Govan, C. D. Meyer, and R. Albright. Generalizing Google’s PageRank to rank national
football league teams. In SAS Global Forum 2008. 2008. Cited on pages 14 and 15.
Z. Gyöngyi, H. Garcia-Molina, and J. Pedersen. Combating web spam with TrustRank . In
Proceedings of the 30th International Very Large Database Conference. 2004. Cited on pages 7
and 22.
T. H. Haveliwala. Topic-sensitive PageRank . In Proceedings of the 11th international conference
on the World Wide Web. 2002. Cited on page 19.
D. J. Higham. Google PageRank as mean playing time for pinball on the reverse web. Applied
Mathematics Letters, 18 (12), pp. 1359 – 1362, 2005. doi:10.1016/j.aml.2005.02.020. Cited
on page 1.
R. A. Horn and S. Serra-Capizzano. A general setting for the parametric Google matrix . Internet
Mathematics, 3 (4), pp. 385–411, 2007. Cited on page 27.
A. Hotho, R. Jäschke, C. Schmitz, and G. Stumme. Information retrieval in folksonomies: Search
and ranking. In Proceedings of the 3rd European Semantic Web Conference, pp. 411–426. 2006.
doi:10.1007/11762256_31. Cited on pages 16 and 19.
B. A. Huberman, P. L. T. Pirolli, J. E. Pitkow, and R. M. Lukose. Strong regularities in World
Wide Web surfing. Science, 280 (5360), pp. 95–97, 1998. doi:10.1126/science.280.5360.95.
Cited on page 24.
A. Jain and P. Pantel. FactRank: Random walks on a web of facts. In Proceedings of the 23rd
34 D. F. Gleich

International Conference on Computational Linguistics, pp. 501–509. 2010. Cited on page 18.
A. Java. Twitter social network analysis. UMBC ebquity blog, http://ebiquity.umbc.edu/
blogger/2007/04/19/twitter-social-network-analysis/, 2007. Cited on page 21.
A. Java, P. Kolari, T. Finin, , and T. Oates. Modeling the spread of influence on the blogosphere.
Technical Report UMBC TR-CS-06-03, University of Maryland, Baltimore, 2006. Cited on page
21.
G. Jeh and J. Widom. SimRank: a measure of structural-context similarity. In KDD ’02: Proceedings
of the eighth ACM SIGKDD international conference on Knowledge discovery and data mining,
pp. 538–543. 2002. doi:10.1145/775047.775126. Cited on pages 11 and 30.
K. Jezek, D. Fiala, and J. Steinberger. Exploration and evaluation of citation networks. In
Proceedings of the 12th International Conference on Electronic Publishing, pp. 351–362. 2008.
Cited on page 17.
B. Jiang. Ranking spaces for predicting human movement in an urban environment. Int. J. Geogr.
Inf. Sci., 23 (7), pp. 823–837, 2009. doi:10.1080/13658810802022822. Cited on pages 8, 13,
and 30.
B. Jiang, S. Zhao, and J. Yin. Self-organized natural roads for predicting traffic flow: a sensitivity
study. Journal of Statistical Mechanics: Theory and Experiment, 2008 (07), p. P07008, 2008.
doi:10.1088/1742-5468/2008/07/P07008. Cited on page 13.
B.-B. Jiang, J.-G. Wang, J.-F. Xiao, and Y. Wang. Gene prioritization for type 2 diabetes in
tissue-specic protein interaction networks. In Third International Symposium on Optimization
and Systems Biology. 2009. Cited on page 10.
Y. Jing and S. Baluja. VisualRank: Applying pagerank to large-scale image search. Pattern
Analysis and Machine Intelligence, IEEE Transactions on, 30 (11), pp. 1877–1890, 2008. doi:
10.1109/TPAMI.2008.121. Cited on page 23.
M. Jockers. Computing and visualizing the 19th-century literary genome. In Digital Humanities.
2012. Cited on page 15.
S. Kamvar. Numerical Algorithms for Personalized Search in Self-organizing Information Networks,
Princeton University Press, 2010. Cited on page 10.
S. Kamvar, T. Haveliwala, C. Manning, and G. Golub. Exploiting the block structure of the web
for computing PageRank . Technical Report 2003-17, Stanford InfoLab, 2003. Cited on pages 10
and 22.
L. Katz. A new status index derived from sociometric analysis. Psychometrika, 18 (1), pp. 39–43,
1953. doi:10.1007/BF02289026. Cited on pages 12, 21, and 25.
J. P. Keener. The Perron-Frobenius theorem and the ranking of football teams. SIAM Review,
35 (1), pp. 80–93, 1993. doi:10.1137/1035004. Cited on page 14.
D. Kempe, J. Kleinberg, and E. Tardos. Maximizing the spread of influence through a social
network . In Proceedings of the ninth ACM SIGKDD international conference on Knowledge
discovery and data mining, pp. 137–146. 2003. doi:10.1145/956750.956769. Cited on page 21.
———. Influential nodes in a diffusion model for social networks. In Proceedings of the 32nd
International Conference on Automata, Languages and Programming, pp. 1127–1138. 2005.
doi:10.1007/11523468_91. Cited on page 21.
M. Kim, R. Sumbaly, and S. Shah. Root cause detection in a service-oriented architecture. In
Proceedings of the ACM SIGMETRICS/international conference on Measurement and modeling
of computer systems, pp. 93–104. 2013. doi:10.1145/2465529.2465753. Cited on page 12.
J. M. Kleinberg. Authoritative sources in a hyperlinked environment. Journal of the ACM, 46 (5),
pp. 604–632, 1999. Cited on page 21.
T. G. Kolda and M. J. Procopio. Generalized BadRank with graduated trust. Technical Report
SAND2009-6670, Sandia National Laboratories, 2009. Cited on page 22.
G. Kollias, E. Gallopoulos, and A. Grama. Surfing the network for ranking by multidamping.
Knowledge and Data Engineering, IEEE Transactions on, PP (99), pp. 1–1, 2013. doi:
10.1109/TKDE.2013.15. Cited on page 24.
G. Kollias, S. Mohammadi, and A. Grama. Network similarity decomposition (NSD): A fast and
scalable approach to network alignment. Knowledge and Data Engineering, IEEE Transactions
on, PP (99), p. 1, 2011. doi:10.1109/TKDE.2011.174. Cited on page 11.
R. I. Kondor and J. D. Lafferty. Diffusion kernels on graphs and other discrete input spaces.
In Proceedings of the Nineteenth International Conference on Machine Learning, pp. 315–322.
2002. Cited on page 24.
E.-M. Kontopoulou, M. Predari, T. Kostakis, and E. Gallopoulos. Graph and matrix metrics
to analyze ergodic literature for children. In Proceedings of the 23rd ACM conference on
Hypertext and social media, pp. 133–142. 2012. doi:10.1145/2309996.2310018. Cited on page
15.
D. Koschützki, K. A. Lehmann, L. Peeters, S. Richter, D. Tenfelde-Podehl, , and O. Zlo-
 PAGERANK BEYOND THE WEB 35

towski. Centrality indicies. In Network Analysis: Methodological Foundations, chapter 3, pp.

16–61. Springer, 2005. doi:10.1007/b106453. Cited on page 1.
J. Kunegis and A. Lommatzsch. Learning spectral graph transformations for link prediction. In
Proceedings of the 26th Annual International Conference on Machine Learning, pp. 561–568.
2009. doi:10.1145/1553374.1553447. Cited on page 24.
H. Kwak, C. Lee, H. Park, and S. Moon. What is Twitter, a social network or a news media? In
WWW ’10: Proceedings of the 19th international conference on World wide web, pp. 591–600.
2010. doi:10.1145/1772690.1772751. Cited on page 21.
A. N. Langville and C. D. Meyer. Deeper inside PageRank . Internet Mathematics, 1 (3), pp.
335–380, 2004. Cited on page 7.
———. Google’s PageRank and Beyond: The Science of Search Engine Rankings, Princeton
University Press, 2006. Cited on pages 1 and 2.
———. Who’s #1? The Science of Rating and Ranking, Princeton University Press, 2012. Cited
on pages 14 and 29.
C. P. Lee, G. H. Golub, and S. A. Zenios. A two-stage algorithm for computing PageRank and
multistage generalizations. Internet Mathematics, 4 (4), pp. 299–327, 2007. Cited on page 28.
D. Liben-Nowell and J. Kleinberg. The link-prediction problem for social networks. Journal
of the American Society of Information Science and Technology, 58 (7), pp. 1019–1031, 2006.
doi:10.1002/asi.20591. Cited on page 20.
X. Liu, J. Bollen, M. L. Nelson, and H. Van de Sompel. Co-authorship networks in the
digital library research community. Inf. Process. Manage., 41 (6), pp. 1462–1480, 2005. doi:
10.1016/j.ipm.2005.03.012. Cited on page 17.
Y. Liu, B. Gao, T.-Y. Liu, Y. Zhang, Z. Ma, S. He, and H. Li. BrowseRank: letting web users
vote for page importance. In SIGIR ’08: Proceedings of the 31st annual international ACM
SIGIR conference on Research and development in information retrieval, pp. 451–458. 2008.
doi:10.1145/1390334.1390412. Cited on page 30.
L. Lü, Y.-C. Zhang, C. H. Yeung, and T. Zhou. Leaders in social networks, the Delicious case.
PLoS ONE, 6 (6), p. e21202, 2011. doi:10.1371/journal.pone.0021202. Cited on page 28.
N. Ma, J. Guan, and Y. Zhao. Bringing PageRank to the citation analysis. Information Pro-
cessing & Management, 44 (2), pp. 800–810, 2008. ¡ce:title¿Evaluating Exploratory Search
Systems¡/ce:title¿ ¡ce:title¿Digital Libraries in the Context of Users Broader Activities¡/ce:title¿.
doi:http://dx.doi.org/10.1016/j.ipm.2007.06.006. Cited on page 17.
M. W. Mahoney, L. Orecchia, and N. K. Vishnoi. A local spectral method for graphs: With
applications to improving graph partitions and exploring data graphs locally. Journal of Machine
Learning Research, 13, p. 23392365, 2012. Cited on pages 25 and 27.
X. Meng. Computing BookRank via social cataloging. 2009. http://cads.stanford.edu/projects/
presentations/2009visit/bookrank.pdf. Cited on page 16.
C. D. Meyer. Matrix analysis and applied linear algebra, Society for Industrial and Applied
Mathematics, Philadelphia, PA, USA, 2000. Cited on page 25.
J. C. Miller, G. Rae, F. Schaefer, L. A. Ward, T. LoFaro, and A. Farahat. Modifications
of Kleinberg’s HITS algorithm using matrix exponentiation and web log records. In SIGIR
’01: Proceedings of the 24th annual international ACM SIGIR conference on Research and
development in information retrieval, pp. 444–445. 2001. doi:10.1145/383952.384086. Cited
on page 24.
B. L. Mooney, L. R. Corrales, and A. E. Clark. Molecularnetworks: An integrated graph
theoretic and data mining tool to explore solvent organization in molecular simulation. Journal
of Computational Chemistry, 33 (8), pp. 853–860, 2012. doi:10.1002/jcc.22917. Cited on
page 10.
J. L. Morrison, R. Breitling, D. J. Higham, and D. R. Gilbert. GeneRank: using search engine
technology for the analysis of microarray experiments. BMC Bioinformatics, 6 (1), p. 233, 2005.
doi:10.1186/1471-2105-6-233. Cited on pages 1 and 10.
M. A. Najork, H. Zaragoza, and M. J. Taylor. HITS on the web: How does it compare?
In Proceedings of the 30th annual international ACM SIGIR conference on Research and
Development in information retrieval (SIGIR2007), pp. 471–478. 2007. doi:10.1145/1277741.
1277823. Cited on page 21.
Z. Nie, Y. Zhang, J.-R. Wen, and W.-Y. Ma. Object-level ranking: Bringing order to web objects.
In Proceedings of the 14th International Conference on World Wide Web, pp. 567–574. 2005.
doi:10.1145/1060745.1060828. Cited on page 18.
G. Osipenko. Dynamical systems, graphs, and algorithms, Springer, 2007. doi:10.1007/
3-540-35593-6. Cited on page 14.
L. Page, S. Brin, R. Motwani, and T. Winograd. The PageRank citation ranking: Bringing order
to the web. Technical Report 1999-66, Stanford University, 1999. Cited on pages 1 and 3.
36 D. F. Gleich

J.-Y. Pan, H.-J. Yang, C. Faloutsos, and P. Duygulu. Automatic multimedia cross-modal
correlation discovery. In KDD ’04: Proceedings of the tenth ACM SIGKDD international
conference on Knowledge discovery and data mining, pp. 653–658. 2004. doi:10.1145/1014052.
1014135. Cited on pages 1 and 18.
G. Pinski and F. Narin. Citation influence for journal aggregates of scientific publications: Theory,
with application to the literature of physics. Information Processing & Management, 12 (5), pp.
297–312, 1976. doi:10.1016/0306-4573(76)90048-0. Cited on pages 16 and 25.
A. Pothen, H. D. Simon, and K.-P. Liou. Partitioning sparse matrices with eigenvectors of graphs.
SIAM J. Matrix Anal. Appl., 11, pp. 430–452, 1990. doi:10.1137/0611030. Cited on pages 25
and 26.
F. Radicchi. Who is the best player ever? a complex network analysis of the history of professional
tennis. PLoS ONE, 6 (2), p. e17249, 2011. doi:10.1371/journal.pone.0017249. Cited on page
15.
S. E. Schaeffer. Graph clustering. Computer Science Review, 1 (1), pp. 27–64, 2007. doi:
10.1016/j.cosrev.2007.05.001. Cited on page 26.
A. Schlote, E. Crisostomi, S. Kirkland, and R. Shorten. Traffic modelling framework for
electric vehicles. International Journal of Control, 85 (7), pp. 880–897, 2012. doi:10.1080/
00207179.2012.668716. Cited on pages 13 and 28.
S. Serra-Capizzano. Jordan canonical form of the Google matrix: A potential contribution to
the PageRank computation. SIAM Journal on Matrix Analysis and Applications, 27 (2), pp.
305–312, 2005. doi:10.1137/S0895479804441407. Cited on page 24.
D. L. Shepelyansky and O. V. Zhirov. Google matrix, dynamical attractors, and ulam networks.
Phys. Rev. E, 81 (3), p. 036213, 2010. doi:10.1103/PhysRevE.81.036213. Cited on page 14.
R. Singh, J. Xu, and B. Berger. Pairwise global alignment of protein interaction networks by
matching neighborhood topology. In Proceedings of the 11th Annual International Conference
on Research in Computational Molecular Biology (RECOMB), pp. 16–31. 2007. doi:10.1007/
978-3-540-71681-5_2. Cited on page 11.
M. Sobek. PR0 - Google’s PageRank 0 penalty. Online., 2003. http://pr.efactory.de/e-pr0.
shtml. Accessed 2013-09-19. Cited on page 22.
Y. Song, D. Zhou, and L.-w. He. Query suggestion by constructing term-transition graphs. In
Proceedings of the Fifth ACM International Conference on Web Search and Data Mining, pp.
353–362. 2012. doi:10.1145/2124295.2124339. Cited on page 20.
O. Sporns. Networks of the Brain, The MIT Press, 2011. Cited on page 11.
W. J. Stewart. Introduction to the Numerical Solution of Markov Chains, Princeton University
Press, 1994. Cited on page 8.
J. J. Sylvester. Chemistry and algebra. Nature, 17, p. 284, 1878. doi:10.1038/017284a0. Cited
on page 10.
J. A. Tomlin. A new paradigm for ranking pages on the world wide web. In Proceedings of the 12th
International Conference on World Wide Web, pp. 350–355. 2003. doi:10.1145/775152.775202.
Cited on page 28.
H. Tong, C. Faloutsos, and J.-Y. Pan. Fast random walk with restart and its applications. In
ICDM ’06: Proceedings of the Sixth International Conference on Data Mining, pp. 613–622.
2006. doi:10.1109/ICDM.2006.70. Cited on page 19.
S. Vigna. TruRank: Taking PageRank to the limit. In WWW ’05: Special interest tracks
and posters of the 14th international conference on World Wide Web, pp. 976–977. 2005.
doi:10.1145/1062745.1062826. Cited on page 25.
———. Spectral ranking. arXiv, cs.IR, p. 0912.0238, 2009. Cited on pages 1, 21, 25, and 27.
K. Voevodski, S.-H. Teng, and Y. Xia. Spectral affinity in protein networks. BMC Systems
Biology, 3 (1), p. 112, 2009. doi:10.1186/1752-0509-3-112. Cited on page 11.
D. Walker, H. Xie, K.-K. Yan, and S. Maslov. Ranking scientific publications using a model of
network traffic. Journal of Statistical Mechanics: Theory and Experiment, 2007 (06), p. P06010,
2007. doi:10.1088/1742-5468/2007/06/P06010. Cited on page 16.
W. Y. Wang, K. Mazaitis, and W. W. Cohen. Programming with personalized PageRank: A
locally groundable first-order probabilistic logic. In Proceedings of the 22Nd ACM International
Conference on Conference on Information and Knowledge Management, pp. 2129–2138. 2013.
doi:10.1145/2505515.2505573. Cited on page 20.
J. Weng, E.-P. Lim, J. Jiang, and Q. He. TwitterRank: finding topic-sensitive influential twitterers.
In WSDM ’10: Proceedings of the third ACM international conference on Web search and data
mining, pp. 261–270. 2010. doi:10.1145/1718487.1718520. Cited on page 21.
J. D. West, T. C. Bergstrom, and C. T. Bergstrom. The Eigenfactor metrics: A network
approach to assessing scholarly journals. College & Research Libraries, 71 (3), pp. 236–244,
2010. arXiv:http://crl.acrl.org/content/71/3/236.full.pdf+html. Cited on page 16.
 PAGERANK BEYOND THE WEB 37

C. Winter, G. Kristiansen, S. Kersting, J. Roy, D. Aust, T. Knösel, P. Rmmele, B. Jahnke,

V. Hentrich, F. Rückert, M. Niedergethmann, W. Weichert, M. Bahra, H. J. Schlitt,
U. Settmacher, H. Friess, M. Büchler, H.-D. Saeger, M. Schroeder, C. Pilarsky, and
R. Grtzmann. Google goes cancer: Improving outcome prediction for cancer patients by
network-based ranking of marker genes. PLoS Comput Biol, 8 (5), p. e1002511, 2012. doi:
10.1371/journal.pcbi.1002511. Cited on page 10.
A. D. Wissner-Gross. Preparation of topical reading lists from the link structure of Wikipedia. In
ICALT ’06: Proceedings of the Sixth IEEE International Conference on Advanced Learning
Technologies, pp. 825–829. 2006. Cited on page 23.
W. Xing and A. Ghorbani. Weighted PageRank algorithm. In Communication Networks and
Services Research, 2004. Proceedings. Second Annual Conference on, pp. 305–314. 2004. doi:
10.1109/DNSR.2004.1344743. Cited on page 8.
H. Yang, I. King, and M. R. Lyu. DiffusionRank: a possible penicillin for web spamming.
In Proceedings of the 30th annual international ACM SIGIR conference on Research and
development in information retrieval, pp. 431–438. 2007. doi:10.1145/1277741.1277815. Cited
on page 24.
A. O. Zhirov, O. V. Zhirov, and D. L. Shepelyansky. Two-dimensional ranking of Wikipedia
articles. The European Physical Journal B, 77 (4), pp. 523–531, 2010. doi:10.1140/epjb/
e2010-10500-7. Cited on pages 13 and 23.
D. Zhou, O. Bousquet, T. N. Lal, J. Weston, and B. Schölkopf. Learning with local and global
consistency. In NIPS. 2003. Cited on page 18.
D. Zhou, J. Huang, and B. Schölkopf. Learning from labeled and unlabeled data on a directed
graph. In ICML ’05: Proceedings of the 22nd International Conference on Machine Learning,
pp. 1036–1043. 2005. doi:10.1145/1102351.1102482. Cited on page 18.
X.-N. Zuo, R. Ehmke, M. Mennes, D. Imperati, F. X. Castellanos, O. Sporns, and M. P.
Milham. Network centrality in the human functional connectome. Cerebral Cortex, 2011.
doi:10.1093/cercor/bhr269. Cited on page 11.

Acknowledgments. We acknowledge the following individuals for their discus-

sions about this manuscript: Sebastiano Vigna, Amy Langville, Michael Saunders,
Chen Greif, Des Higham, and Stratis Gallopoulos, as well as Kyle Kloster for carefully
reading several early drafts. This work was supported in part by NSF CAREER award
CCF-1149756.