Typesetting math: 97%

Algorithmic cheatsheet
2014-10-02

General results

Data structures

Sorting

Graph algorithms

This page sums up some important results from computer science. They are extracted from theIntroduction to Algorithms (Third Edition), by
Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest and Clifford Stein. We highly recommend it.
The following information is organized in several sections grouping definitions. Each definition points to the Introduction to Algorithms for
further information using the abbreviation CLRS (from the authors' names).
The formul are written using LATEX and the final render of the page is done by the Javascript library MathJax (currently the simplest way to
use LATEX in HTML).
The current figures use images from external websites. Clicking on it will redirect you to their original webpages.

General results
Master Theorem
CLRS: p93
Recurrence of the form: T(n)=aT(n/b)+O(f(n))
Case

f(n)=O()

T(n)=O()

nlogba

nlogba

nlogba

lognnlogba

Case

f(n)=O()

T(n)=O()

nlogba+ and c,af(n/b)cf(n)

f(n)

http://anupcowkur.com/posts/master-theorem-simplified/ http://www.geeksforgeeks.org/analysis-algorithm-set-4-master-method-solvingrecurrences/ http://rosalind.info/glossary/algo-master-theorem/

call tree; b defines the maximal depth of the recursive calls, a defines the maximal number of calls to f at each level, summing the complexity
of each of the O(nlogba) calls to f, which depends on the depth, explains the cases of the master theorem (the illustraction come from the CLRS
Second Edition)

Matroid
CLRS: p437
Pair (S,I) with:

S is a finite set

IP(S) (independent subsets)

ABI,AI (heredity)

A,BI,|A|<|B|xBA,A{x}I(exchange)

Greedy algorithms from matroids

CLRS: p439

(S,I) is a matroid. w is a weight function over S. Greedy algorithm choosing iteratively an element maximizing w:

{x}I with maximum weight

x is in any independent subset with maximum weight

I={BS{x}:B{x}I} (subsets containing x, ignoring x)

S={yS,{x,y}I) (restricting elements)

iterate on (S,I) (also a matroid)

Data structures
Heap
CLRS: p151
Sorted data structure (priority queue) Min-heap: parent is bigger than its children Max-heap: parent is bigger than its children

Speed
we do have 84 > 38,29 and 38 > 18,17
Operation

Complexity

Memory use

O(n)

Insert

O(logn)

Maximum

O(1)

Extract-Max

O(logn)

Increase-Key

O(logn)

Hash Table
CLRS: p256
Arbitrarily indexed table h is a hash function to [1..m]

Space: O(m)

Locate: O(1+n/m)

Note: Locate finds a location in the array where the user can read or write as usual.

Binary Search Tree (BST)

CLRS: p286
Sorted data structure h is the height of the tree

Space: O(n)

Search: O(h)

Insert: O(h)

Delete: O(h)

Red-Black Tree
CLRS: p308
BST maintaining h=logn with:

every node is either black (normal) or red (to be updated);

the root is black;

leaves are black;

red nodes cannot have a red parent there are no more red nodes than black ones;

all simple paths from the root to a leaf have the same number of black nodes, limiting h.
any simple path from the root to a leaf has 3 black nodes, there are no more red nodes than black ones. (source)

B-Tree
CLRS: p484
Variant of the BST using a large fan-out t to reduce disk-accesses Search, insert and delete takeO(\log_t n) disk accesses and O(t \log_t n) CPU
time.
the fan-out is usually around 1000, making two levels enough most of the time (source)

Union-find
CLRS: p568
Data structure for sets, supporting union of sets and finding the representative m is the number of previously run Make-Set operations. \alpha is
the Ackerman function (\alpha(m) \leq 4 in practise).

Make-Set: O(1)

Union: O(1)

Find-Set: O(\alpha(m))

http://scienceblogs.com/goodmath/2007/09/06/graph-searches-and-disjoint-se/

Sorting
Bubblesort
CLRS: p40 (problem 2-2) Complexity: O(n^2)
Fixing inverted adjacent values n times

the red squares highlight the currently checked pairs, all pairs have to be checked n times (source)

Insertion sort
CLRS: p16 (correctness), p24 (complexity) Complexity: O(n^2)

Inserting each element in a new array at the right location

Merge sort
CLRS: p29 Complexity: O(n \log n)
Divide-and-conquer (sort each half, then merge)

Heapsort
CLRS: p159 Complexity: O(n \log n)
Building a heap on the n elements and extracting them all

Quicksort
CLRS: p170 Complexity: O(n \log n)
Divide-and-conquer (choose a pivot, then filter lower and greater values)

the elements are split in two sub-arrays using pivot as the limit value and each part is sorted
recursively (source)

Graph algorithms
{Breadth,Depth}-first search
CLRS: p594 (breadth), p603 (depth) Complexity: O(E + V)
Listing the vertices of a graph. Breadth-first search list the siblings, then the siblings' siblings
visit order of a breadth-first-search algorithm (source)

Depth-first search goes as far as it can before considering siblings.

visit order of a depth-first-search algorithm (source)

Kruskal
CLRS: p631 Complexity: O(E \log V)
Finding a minimum spanning tree Consider every edge in nondecreasing order, add to current forest if links two components. The components
are handled using union-find.

Prim
CLRS: p634 Complexity: O(E \log V) (binary heap) Complexity: O(E + V \log V) (Fibonacci heap)
Finding a minimum spanning tree Iteratively extract closest vertex u and relax all edges (u, v). The set of vertices is handled using a priority
queue.

Bellman-Ford
CLRS: p651 Complexity: O(E V)
Finding all shortest paths from a single source Just relax every edge of the graph |V| times Detects negative weight cycles

Shortest paths from DAG

CLRS: p655 Complexity: O(V+E) (including sorting)

Finding all shortest paths from a single source in Directed Acyclic Graphs Consider every vertex uin topological order and relax all edges (u, v)

Dijkstra
CLRS: p658 Complexity: O(V^2 + E) (array) Complexity: O((E+V) \log V) (binary heap)Complexity: O(V \log V + E) (Fibonacci heap)
Finding all shortest paths from a single source with positive weights Iteratively extract closest vertex u and relax all edges (u, v). The set of
vertices is handled using a priority queue.

Floyd-Warshall
CLRS: p693 Complexity: O(V^3)
Finding the shortest path from all vertex pairs Relaxing pair distance |V| times

The Path-Finding Problem

Speed
Diagonals
we want to find a shortest or near-shortest path; Breadth First Search works well on unweighted graph (no water); Dijkstra generalize it to
weighted graph; A* is a common path finding used in games which generalizes Dijkstra and Best First Search
Contact:
[email protected]