Scribd
Upload
7 views

Hashing Practice Problems Hints

DSA qs

Uploaded by

Sneha Dhaka
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
7 views

Hashing Practice Problems Hints

DSA qs

Uploaded by

Sneha Dhaka
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Practice Sheet I: Hashing

COL106: Data Structures and Algorithms


Semester-I 2023–2024

1 Problem I: Basic Concepts


Given input {4371, 1323, 6173, 4199, 4344, 9679, 1989} and a hash function h(x) = x(mod 10), show the
resulting:
1. Separate chaining hash table
2. Hash table using linear probing
3. Hash table using quadratic probing

4. Hash table with second hash function h2 (x) = 7 − (x mod 7)

Problem II: Needle in a Haystack


Suppose you are given a pattern text called needle of length k and you need to find the first occurrence of
this string in a large text haystack of size n.
1. Think of a simple algorithm to solve this problem. What is its time complexity?
2. Suppose you have an efficient hash function for hashing strings. If you are allowed to have a few false
positives of needle in the haystack, can you think of an approach to solve this problem in a faster
way? Analyse the time complexity of your approach.
3. Prove that the expected number of false positives will be few.
Hints:
1. Brute force approach: Slide the string needle over the string haystack. Time taken = O(nk)

2. Still slide the needle over the haystack, but compare the hash values of the key k with the hash
values of the window you are checking against. For this to be efficient, you need your hash function to
be fast, such that you can calculate the next hash from the previous hash in O(1) time. This is called
a rolling hash.

3. The probability of false positives is equal to the probability of collisions of the hash function you choose.
So, if the hash function is good, the probability will be low.

Problem III: Hashing and Probability


Let U = [1, M ] be a universe, and let S = {s1 , · · · , sn } be a subset of U of size n such that each si is
a uniformly random element of U independent of other sj ’s. We are implementing a hash table T with
chaining. T [i] represents the ith chain. Let H be a hash function such that H(x) = x mod n.

1
n−1
1. Show that the expected size of maxi=0 T [i] is O(logn).
Note: The expected value of any random variable X is defined as,
X
E(X) = xP (X = x)
x

2. Argue that the expected value of maximum time taken to verify the membership of elements of U in
S is O(logn)
Hints:
Let X = maxi T [i]. Note we just want an upper bound on E(X). Here are some questions for you to think
about:

• Consider the event Ei , which is that the size of the ith chain is greater than some constant k. Can you
find an upper bound on P (Ei )?
• Consider the event E = ∪i Ei . What does this event signify? Can you find (an upper bound on) P (E)?
• Since the maximum value of X is n, convince yourself that you can bound the E(X) term as the
following:
E(X) ≤ nP (E) + kP (E c )

• Choose an appropriate value of k and complete the proof.

Problem IV: Amortized Analysis of Open Addressing


Consider a simple open addressing scheme, let’s say linear probing with a hash code f0 (x). We start with
an array of size n, use hash function fn (x) = f0 (x) mod n. When this array gets full we move to an array
of size 2n with hash function f2n (x) = f0 (x) mod 2n. When this gets full we again double the size and so
on. Clearly a single insert could take a long time if rehashing is to be done. Show that the amortized insert
time is θ(1).

Hints: Use an accounting method over a series of hashtable operations, similar to the analysis of stacks
using dynamically sized arrays. You may want to refer to the analysis presented here.

Problem V: Cubic Probing


Suppose instead of quadratic probing, we use “cubic probing”; here the ith probe is at hash(x) + i3 . Do you
think cubic probing improves on quadratic probing?

Hints: The issues with both quadratic and cubic probing is that there is secondary clustering. A cubic
will grow faster, but therefore will also get modded out faster (since the size of the hashtable is the same).
There should not be any practical improvement in general.

You might also like