For any Homework related queries, Call us at : -  

+1 678 648 4277

You can mail us at : - [email protected] or
reach us at : - https://www.programminghomeworkhelp.com/

Algorithms Homework Help

Problem 2.
The following two Python functions correctly solve the problem:
given an array X of n positive integers, where the maximum integer
in X is k, return the integer that appears the most times in X.
Assume: a Python list is implemented using a dynamic array; a
Python dict is implemented using a hash table which randomly
chooses hash functions from a universal hash family; and max(X)
returns the maximum integer in array X in worst-case O(|X|) time.
For each function, state its worst-case and expected running times in
terms of n and k.

(a) (b)
def frequentest_a(X): def frequentest_b(X):
k = max(X) k = max(X)
H = {} A = []
for x in X: for i in range(k + 1):
H[x] = 0 A.append(0)
best = X[0] best = X[0]
for x in X: for x in X:
H[x] += 1 A[x] += 1
if H[x] > H[best]: if A[x] > A[best]:
best = x best = x
return best return best
i. Worst-case: i. Worst-case

Solution: O(n + k)
Solution: O(n2)

ii. Expected: i. Expected:

Solution: O(n) Solution: O(n + k)

Common Mistakes: Saying doing a O(1)-time operation m

times takes amortized O(m) time.
Problem 3. Haphazard Heap

Array [A,B,C,D,E,F,G,H,I,J] represents a binary min-heap

containing 10 items, where the key of each item is a distinct integer.
State which item(s) in the array could have the key with:

(a) the smallest integer

Solution: A

(b) the third smallest integer

Solution: B, C, D, E, F, G

Common Mistakes: Incorrectly assuming third smallest must be in

either B or C.

(c) the largest integer

Solution: F,
G, H, I, J
Common Mistakes: Thinking the largest integer must be in the
bottom level (instead of in any leaf)

Problem 4. Transforming Trees

The tree below contains 8 items, where each stored item is an

integer which is its own key.

a) Suppose the tree drawn above is the implicit tree of a binary max-
heap H. State the array representation of H, first before and then
after performing the operation H.delete max().
 Solution:

Before: [93, 85, 38, 23, 63, 11, 13, 17]

After: [85, 63, 38, 23, 17, 11, 13]

Common Mistakes:

• Not building heap correctly (e.g., confusing heap order with

traversal order)
• Not correctly deleting min (by swapping root with last leaf and
heapifying down)
(b) Suppose instead that the original tree drawn above is a
Sequence AVL Tree S (note Sequence data structures are zero-
indexed). The items in the leaves of S in traversal order are (17,
63, 11, 13). Perform operation S.delete at(3) on S including any
rotations, and then list the items stored in the leaves of S in
traversal order, after the operation has completed. (You do not
need to draw the tree.)
Solution: (17, 85, 11, 13)
Common Mistakes:

• Indexing by heap order instead of traversal order (or using 1-

indexing)
• Listing full traversal order (rather than just the leaves)
• Deleting the item but not rebalancing the tree to satisfy the AVL
Property
Problem 5. Sorting Sock

At Wog Hearts School of Wizcraft and Witcherdry, n incoming

students are sorted into four houses by an ancient magical
artifact called the Sorting Sock. The Sorting Sock first sorts the n
students by each of the four houses’ attributes and then uses the
results to make its determinations. For each of the following
parts, state and justify what type of sort would be most efficient.
(By “efficient”, we mean that faster correct algorithms will
receive more points than slower ones.)

(a) For House Puffle Huff, students must be sorted by friend

number, i.e., how many of the other n - 1 incoming students
they are friends with, which can be determined in O(1) time.

Solution: Friend numbers are non-negative integers less than n,

so we can use counting sort to sort the students in worst-case
O(n) time. (Radix sort also works with the same running time.)
Since we have to compute friend number for each student, any
algorithm will take at least Ω(n) time, so this is optimal.

Common Mistakes: Using a sort that is not O(n)

(b) For House Craven Law, students must be sorted by the weight
of their books. Book weights cannot be measured precisely, but the
Sorting Sock has a scale that can determine in O(1) time whether
one set of books has total weight greater than, less than, or equal
to another set of books.

Solution: A scale weighing is a comparison with a constant

number of outcomes, so the comparison sort Ω(n log n) lower
bound applies. So we cannot do better than by using an worst-case
O(n log n) sorting algorithm, e.g., merge sort, using the scale to
compare one student’s books against another’s.

Common Mistakes: Insufficient justification for why O(n log n)

is optimal
(c) For House Driven Gore, students must be sorted by bravery,
which can’t be directly measured or quantified, but for any set of
students, the Sorting Sock can determine the bravest among
them in O(1) time, e.g., by presenting the students with a scary
situation.
Solution: We can’t quantify bravery, so we can’t hope to use any
integer-based algorithms. However, the Sorting Sock can find a
student of maximum bravery in O(1) time, so we repeatedly find
and select a bravest student among all previously unselected
students in worst-case O(n) time, which is again optimal. (This is
priority queue sort, using the Sorting Sock as the priority queue to
find the maximum.)
Common Mistakes: Arguing a Ω(n log n) lower bound

(d) For House Leather Skin, students must be sorted by their

magical lineage: how many of a student’s ancestors within the
previous 3dlog ne + 4 generations were magical. Recall that
humans, magical or not, always have two parents in the previous
generation, unlike binary tree nodes which have at most one.
Assume the Sorting Sock can compute the magical lineage of a
student in O(1) time.
Solution: Each student has at most 2k ancestors in the kth
generation preceding. Thus the number of wizard ancestors will be
a non-negative number bounded above

Thus we can use radix sort to sort the students by their magical
lineage in worst-case O(n) time, which is again optimal.

Common Mistakes:
• Claiming ancestors instead of
• Saying that and using counting sort
Problem 6. Triple Sum

Given three arrays A, B, C, each containing n integers, give an

O(n2)-time algorithm to find whether some a ∈ A, some b ∈ B, and
some c ∈ C have zero sum, i.e., a + b + c = 0. State whether your
running time is worst-case, expected, and/or amortized.

Solution: We present both expected and worst-case solutions.

Expected Time

For each pair of numbers (a, b) ∈ A × B, store a + b in a hash table

H. Then return Yes if -c appears in H for any c ∈ C, and return No
otherwise.
Correctness: If any-c appears in H for any c ∈ C, then -c = a0 + b0
for some (a0 , b0 ) ∈ A × B so a0 + b0 + c = 0. Otherwise, there is
no c ∈ C for which -c = a0 + b0 for any (a0 , b0 ) ∈ A × B.
Running Time: There are O(n2) pairs in A × B, so inserting them
into H takes expected O(n2) time. Then checking whether each -c
appears in H takes expected O(1) time each, and expected O(n) in
total. So this algorithm runs in expected O(n2) time.
Continued on scratch paper S1 for worst-case solution...

Common Mistakes:
• Using counting/radix sort or creating a direct access array (no
bound on u so not efficient)
• Saying n insertions into a hash table gives an amortized bound
• Checking all triples in Ω(n3) time
 Problem 7. Where Am i?

Given a Sequence AVL Tree T containing n nodes, and a pointer to

a node v from T , describe an O(log n)-time algorithm to return the
(zero-indexed) index i of node v in the traversal order of T. (Recall
that every node u in a Sequence AVL Tree T stores an item u.item,
parent u.parent, left child u.left, right child u.right, subtree height
u.height, and subtree size u.size.)

Solution: Our algorithm will be to walk up the tree from v to the root
r of the Sequence AVL Tree, counting the nodes preceding v in the
traversal order along the way, since the number of nodes preceding v
in the tree is equivalent to v’s (zero-indexed) index.

Let #v(u) be the number of vertices preceding v in a vertex u’s

subtree, where v is in the subtree of u. Then #v(v) = v.left.size if v
has a left child and zero otherwise; and can be computed in O(1)
time. Then, for every ancestor of u starting from v, we compute
#v(u.parent) from #v(u). There are two cases:
• Case 1, u is the left child of u.parent: then all the nodes preceding
v in the subtree of u.parent are in the subtree of u, so set
#v(u.parent) = #v(u).

• Case 2, u is the right child of u.parent: then all nodes in the left
subtree of u.parent precede v (as does u), so set #v(u.parent) = 1 +
u.parent.left.size + #v(u).
Then return #v(r), since this is the number of nodes preceding v in
r’s subtree (i.e., the entire tree). Correctness is argued within the
algorithm description. This algorithm spends worst-case O(1) work
for each ancestor of v, so since the Sequence AVL Tree is balanced,
the number of ancestors is bounded by O(log n), and the algorithm
runs in worst-case O(log n) time.

Common Mistakes:

• Finding v given i instead of finding i given v

• Breaking early (e.g., as soon as node is a left child, instead of
continuing up tree) • Walking down the tree from root (assuming
the way to go to find v)

Problem 8. Methane Menace

FearBird is a supervillian who has been making small holes in the

methane gas pipe network of mahtoG City. The network consists
of n pipes, each labeled with a distinct positive integer. A hole i is
designated by a pair of positive integers (pi, di), where pi denotes
the label of the pipe containing the hole, and di is a positive integer
representing the distance of the hole from the front of pipe pi.
Assume any two holes in the same pipe pi will be at different
distances from the front of pi. When a new hole (pi, di) is spotted,
the city receives a report of the hole to keep track of. The city will
periodically patch holes using the following priority scheme:
• if each pipe contains at most one hole, patch any hole (if one exists);

• otherwise, among pairs of holes (pi, di) and (pj , dj ) appearing

on the same pipe, i.e., pi = pj , identify any pair with smallest
distance |di - dj | between them, and patch one of them.

Describe a database supporting the following operations, where k

is the number of recorded but unpatched holes in the network at
the time of the operation. State whether your running times are
worst-case, expected, and/or amortized.

initialize(H) Initialize the database with n holes H =

{(p0, d0), . . . ,(pn-1, dn-1)}, with one hole
on each pipe, in O(n) time

report(pi,di) Record existence of a hole in pipe pi at

distance di in O(log k) time

patch() Patch any hole that follows the priority

scheme above in O(log k) time

Solution: To implement the database, maintain the following data

structures:
• A Set AVL tree Tp for each pipe p containing all the unpatched
holes in p keyed by hole distance
• A Hash Table D mapping each pipe p to its tree Tp
• A Binary Min Heap Q containing each consecutive pair of holes (p,
d1, d2) appearing on the same pipe p with key being the distance |d2
-d1| between them, and any lonely holes (p, d) (holes that are alone
on their pipes) with key ∞ (when multiple stored items have the same
key, we store them in a Hash Table keyed by (p, d1, d2) or (p, d))
• A Hash Table C mapping each consecutive hole pair (p, d1, d2) or
lonely hole (p, d), to their location in Q.

Some parenthetical notes on this solution:

• A Set AVL Tree can be used for C or Q to achieve identical bounds.

• A solution without augmentation was intended, so our solution does
not use it. But it is also possible to use augmentation:

– e.g., to combine C and Q into a single Set AVL Tree keyed

the same as C but storing a pointer to the min distance in
subtree; or,
– e.g., on each Tp to maintain the min distance within the pipe (to
compute this augmentation efficiently, one would either need to
maintain the distance to each hole’s successor/predecessor (not a
subtree property), or augment by min/max distance in subtree to be
correct)
Operation descriptions: Continued on scratch paper S2...

Problem 9. Vapor Invite

Vapor is an online gaming platform with n users. Each user has a

unique positive integer ID di and an updatable status, which can be
either active or inactive. Every day, Vapor will post online an active
range: a pair of positive integers (a, b) with the property that every
user having an ID di contained in the range (i.e., with a ≤ di ≤ b)
must be active. Vapor wants to post an active range containing as
many active users as possible, and invite them to play in a special
tournament. Describe a database supporting the following worst-
case operations:
 build(D) Initialize the database with user IDs
D = {d0, . . . , dn-1}, setting all user
statuses initially to active, in O(n log
n) time
toggle status(di) Toggle the status of the user with ID di,
e.g., from active to inactive or vice
versa, in O(log n) time
big active range() Return an active range (a, b)
containing the largest number of
active users possible in O(1) time

Solution: To implement the database, maintain a single Set AVL

Tree T containing each user ID and their status, keyed by ID. In
additional, augment each node x in T with four subtree
properties:

• x.size: the number of IDs in the subtree (as discussed in

lecture).
• x.suffix = (d, m): the smallest ID d in the subtree for which
each of the m IDs d0 ≥ d in the subtree is active, or None.
Computable in O(1) time as either the suffix (dR, mR) of the
right subtree, or if mR is equal to the size of the right subtree
and x is active, return the suffix (dL, mL) of the left subtree
but add mR + 1 to mL (or (x.key, mR + 1) if left suffix is
None).
• x.prefix = (d, m): the largest ID d in the subtree for which
each of the m IDs d0 ≤ d in the subtree is active, or None.
Computable in O(1) time as either the prefix (dL, mL) of the
right subtree, or if mL is equal to the size of the left subtree and
x is active, return the prefix (dR, mR) of the right subtree but
add mL+1 to mR (or (x.key, mL+1) if right prefix is None).
• x.substring = (a, b, m): a, b are IDs from the subtree where
each of the m IDs d in the subtree with a ≤ d ≤ b is active and
m is maximized. Computable in O(1) time by taking the max of
the substring within either left or right subtree, or the substring
spanning the two subtrees if x is active. Specifically, consider
the substrings of left and right subtrees, (aL, bL, mL) and (aR,
bR, mR) respectively, and then if x is active, consider the suffix
of the 0 0 0 left subtree (dL, m L + 1 + m0 ) and the prefix of
the right subtree (dR, m R, mR) is the largest, return (aL, bL,
mL), (dL, dR, m0 ). Then depending on which L + 1 + m L R 0
of (mL, m R), or (aR, bR, mR) respectively.

To implement build(D), build the Set AVL Tree T in worst-case

O(n log n) time, maintaining the custom augmentations during
each insertion.
To implement toggle states(di), remove di from T in worst-case
O(log n) time, toggle its status in O(1) time, and then re-insert di
into T in worst-case O(log n) time (again, maintaining
augmentations).
To implement big active range(), simply return the substring
augmentation at the root in worst-case O(1) time, which is
correct by the definition of our augmentation.
Common Mistakes: See scratch paper S3...

You can use this paper to write a longer solution if you run out of
space, but be sure to write “Continued on S1” on the problem
statement’s page.
Solution: (Problem 6 continued...)

Worst-case Solution

Sort A and B increasing using merge sort. We give a two-finger

algorithm to determine whether any a ∈ A and b ∈ B sum to a given
-c. Doing this for each c ∈ C directly determines whether a, b, c sum
to 0.

Start with an index i = 0 into A and an index j = n - 1 into B and

repeat the following procedure until either i = n or j = -1:

• If A[i] + B[j] > -c, increase i.

• If A[i] + B[j] < -c, decrease j.
• Otherwise, A[i] + B[j] = -c, so return Yes.

If this procedure terminates without returning Yes, return No.

Correctness: We first prove the claim that the loop maintains the
invariant that at the start of a loop, that no A[i 0 ] + B[j0 ] = -c for
any 0 ≤ i 0 < i or any n - 1 ≥ j0 > j. Proof by induction on k = i - j.
This invariant is trivially true at the start when k = i - j = -n + 1.
Assume the claim is true for all k < k∗ = i ∗ - j ∗. If we are at the
start of a loop with i = i ∗ and j = j∗, we cannot have yet returned, so
at the start of the previous loop, there are two cases:

• Case 1, (i, j) = (i ∗ - 1, j): i ∗ -1 - j = k∗ - 1 < k∗, so by

induction, no A[i 0 ] + B[j0 ] = -c for any i 0 < i∗ - 1 or j0 > j ∗.
But we increased i when A[i ∗ - 1] + B[j∗] > -c, so since B[j ∗] ≥
B[j0 ] for all j0 > j∗, then A[i ∗ - 1] + B[j0 ] > -c for all j0 > j ∗,
restoring the invariant.

• Case 2, (i, j) = (i ∗, j +1): i ∗-(j +1) = k∗-1 < k∗, so by induction,

no A[i 0 ]+B[j0 ] = -c for any i 0 < i∗ or j0 > j∗ + 1. But we
decreased j when A[i ∗] + B[j∗ + 1] < -c, so since A[i ∗] ≤ A[i 0 ]
for all i 0 < i∗, then A[i 0 ] +B[j0 +1] < -c for all i 0 < i ∗, restoring
the invariant.

So the claim is true. Then if the algorithm terminates without

returning Yes, either i = n or j = -1, so the claim implies that no
integer from either A or B respectively can be in a triple with c that
sums to zero.
 Running Time: Sorting A and B takes worst-case O(n log n) time. We
perform the two-finger algorithm n times. A single two-finger
algorithm takes worst-case O(n): every loop does O(1) work and
either increases i or decreases j, so since the loop terminates when
either i = n or j = -1, the loop executes at most 2n = O(n) times. So
the algorithm runs in worst-case O(n2) time in total.

You can use this paper to write a longer solution if you run out
of space, but be sure to write “Continued on S2” on the problem
statement’s page.

Solution: (Problem 8 continued...)

To implement initialize(H), initialize empty D and C, and then for
each (p, d) ∈ H, construct an empty Set AVL Tree Tp, insert d into
Tp, insert p into D mapping to Tp. Then build a hash table D∞ on
every (p, d) ∈ H, store it in Q with key ∞, and then insert each (p,
d) into C mapping to D∞. Building D∞ takes expected O(n) time,
and for each (p, d), this procedure takes O(1) time (expected in the
case of inserting into D). So it takes expected O(n) time in total,
and maintains the invariants of the database directly. (Note that we
could not use a Set AVL Tree for D, as it could take Ω(n log n)
time to construct.)

To implement report(p, d), lookup p in D to find Tp, and then

insert d into Tp.
• If d has no predecessor or successor in Tp, then insert (p,
d) into Q with key ∞ and insert (p, d) into C mapping to its
location in Q.
• Otherwise d has a predecessor or successor in Tp

– If d has a predecessor d1 and successor d2, lookup (p, d1, d2)

in C to find it in Q, and then remove (p, d1, d2) from both C and
Q.
– Otherwise, it has one of them d0 , so lookup (p, d0 ) in C to
find it in Q and then remove (p, d0 ) from both C and Q.
– In either case, if d has a predecessor d1, add (p, d1, d) to Q
with key |d - d1| and add (p, d1, d) to C pointing to its location in
Q; – and if d has a successor d2, add (p, d, d2) to Q with key |d2
-d| and add (p, d, d2) to C pointing to its location in Q.

This procedure does a constant number of worst-case O(log k)

time or amortized expected O(1) time operations, so this
operation runs in amortized expected O(log k) time, and
maintains the invariants of the database by removing any
consecutive pairs or lonely holes if they are no longer
consecutive or lonely, and adding any new consecutive pairs or
lonely holes that may have been introduced.
To implement patch(), delete the minimum item from Q containing
one or two holes on pipe p, remove it from C, lookup p in D to
find Tp, and remove the relevant holes from Tp. This procedure
does a constant number of worst-case O(log k) time or amortized
expected O(1) time operations, so this operation runs in amortized
expected O(log k) time, and maintains the invariants of the
database directly. It is correct because Q exactly implements the
requested priority scheme: consecutive pairs with smaller key have
higher priority, and will only remove a lonely hole if there are no
consecutive pairs having finite key contained in Q.

Common Mistakes: See scratch page S3...

You can use this paper to write a longer solution if you run out of
space, but be sure to write “Continued on S3” on the problem
statement’s page.

Common Mistakes:
• Taking a minimum in some pipe rather than a minimum over all
pipes
• Not prioritizing pipes having more than one hole over those
with only one
• Storing an AVL or heap on all pipes which cannot be maintained
within the time bounds
• Attempting an augmentation-based solution incorrectly
• Claiming O(log n)-time insertion into a sorted array
• Initializing a direct access array of non-polynomially bounded size
• Using Sequence AVL Trees instead of Set AVL Trees

Common
Mistakes:
• Assuming the max range always goes through the root
• Claiming a substring augmentation without showing how to
maintain it in O(1) time
• Using 2 AVL Trees, one of active and one of inactive (doesn’t help
to find largest range).
• Substring augmentation maintenance doesn’t consider substrings
containing the root