Yanxi Chen

Software Engineer@Amazon

3 minute read

Some Notes of Introduction to Algorithm

Fiboniacci Number

\[\begin{bmatrix} F_{n+1} & F_{n} \\ F_{n} & F_{n-1} \\ \end{bmatrix} = \begin{bmatrix} 1 & 1 \\ 1 & 0 \\ \end{bmatrix}^n\]

$Running\ Time = \theta(\log_2(n))$

Order Statistics

Given n elements in an array, find $k^{th}$ smallest element.

  • Quick Select
    • Expected running time $\theta(n)$
    • Worse case $\theta(n^2)$
  • Worse-case linear time order statistics 
    Select(i, n)
1. Divide n elements into [n/5] groups of 5 elements each. Find the median of each
group. O(n)
2. Recurrsively select the medium x of the [n/5] group medians. T(n/5)
3. Partition with x as pivot, let k = rank(x). O(n)
4. if i==k then return x
  if i<k then recurrsively select ith smallest element in left part
  else then recurrsively select (i-k)th smallest element in upper part

Hash Functions

Division Method

$h(k) = k\ mod\ m$

pick $m$ to be prime and not too close to power of $2$ or $10$.

Multiplication Method

$h(k)$ $=$ $A\cdot k$ $mod$ $2^w$ » $(w - r)$, $A\ odd\land2^{w-1}$ < $A$ < $2^w$

Universal Hashing

Let $u$ be a universe of keys, and let $H$ be a finite colleciton of hash functions mapping $U$ to {$0,1,\dots,m-1$}.

$H$ is $universal$ if $\forall x,y\in U,x\ne y$

\[\lvert\{h\in H;h(x)=h(y)\}\rvert=\lvert H\rvert/m\]

i.e. if $h$ is chosen randomly from $H$, the probability of collision between $x$ and $y$ is $1/m$.

Perfect Hashing

Given $n$ keys, construct a static hash table of size $m=O(n)$ such that searching takes $O(1)$ time in the worst case.

Idea: 2 level scheme with universal hashing at both levels and NO collisions at level 2.

if $n_i$ items that hashes to level 1 slot $i$, then use $m_i=n_i^2$ slots in the level 2 table $S_i$.

Augmented Data Structures

Dynamic Order Statistics

Supports: Insert, Delete, Search(x), Select(i), Rank(x).

Idea: use a R-B tree while keeping sizes of the subtree.

$size[x]=size[left(x)]+size[right(x)]+1$

Select(root, i):
    k = size[left(x)] + 1 // k = rank(x)
    if i == k then return x
    if i < k then return Select(left(x), i)
    else return Select(right(x), i - k)

$Running\ Time = \theta(\log_2(n))$

Interval Tree

Supports: Intert, Delete, Interval-Search: Find an interval in the set that overlaps a given query interval.

Idea: use a R-B tree while keeping the largest value $m$ in the subtree.

\[m[x]=max\{high[int[x]], m[right(x)], m[left(x)]\}\] 
Interval-Search(i) // finds an interval that overlaps i
    x = root
    while x != nil and (low[i] > high[int[x]] or low[int[x]] > high[i]) do // i and int[x] don't overlap
        if left[x] != nil and low[i] <= m[left[x]] then x = left[x]
        else x = right[x]
    return x

Amortized Analysis

Potential Method

Framework:

  • Start with data structure $D_0$
  • operation $i$ transforms $D_{i-1} \to D_i$
  • cost of the operation is $c_i$
  • Define a potential function:
\[\Phi:\{D_i\}\to\mathbb{R}\ such\ that\ \Phi(D_0)=0\land\Phi(D_i)\geq0\forall i\]
  • Amortized cost $\hat{c_i}$ with respect to $\Phi$ is
\[\hat{c_i}=c_i+\Phi(D_i)-\Phi(D_{i-1})\]
  • Total amortized cost of n operations is
\[\begin{align} \sum_{i=1}^{n}\hat{c_i}&=\sum_{i=1}^{n}(\hat{c_i}+\Phi(D_i)-\Phi(D_{i-1}))\\ &=\sum_{i=1}^{n}\hat{c_i}+\Phi(D_n)-\Phi(D_0)\\ &\geq\sum_{i=1}^{n}c_i \end{align}\]

Competitive Analysis

An online algorithm A is $\alpha$-$competitive$ if $\exists k$ such that for any sequence of operations $S$,

\[Cost_A(S)\leq\alpha\cdot C_{opt}(S)+k\]

where $C_{opt}(S)$ is the optimal, off-line, “God’s” algorithm.

Karp-Rabin Algorihm: Find s in t

Rolling Hash ADT:

  • r.append(c): r maintains a string x where $r=h(x)$, add char c to the end of x
  • r.skip(): delete the first char of x. (assume it is c).

Then just use ADT to “roll over” t to find s.

Note: If their hashes are equal, there is still a probability $\leq 1/\lvert S\rvert$ that they are actual not the same string.

To implement ADT: use hash simple hash function $h(k)=k\bmod m$ where $m$ is a random prime $\geq\lvert S\rvert$

We can treat $x$ as a multidigit number $u$ in base $a$, where $a$ is just the alphabet size.

So:

  • $r()=u\bmod m$
  • $r$ stores $u\bmod m$ and $\lvert x\rvert$, (really $a^{\lvert x\rvert}$), not $u$.
r.append(c)
    u = u * a + ord(c) mod m 
      = [(u mod p) * a + ord(c)] mod m
      = [r() * a + ord(c)] mod m

r.skip(c) // assume char c is skipped
    u = [u − ord(c) * (pow(a, |u| - 1) mod p)] mod p
      = [(u mod p) − ord(c) * (pow(a, |u| - 1) mod p)] mod p
      = [r() − ord(c) * (pow(a, |u| - 1) mod p)] mod p

You May Also Enjoy