eleg667/2001-f/topic-1a 1 a brief review of algorithm design and analysis

42
eleg667/2001-f/Topi c-1a 1 A Brief Review of Algorithm Design and Analysis

Post on 20-Dec-2015

223 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 1

A Brief Review of

Algorithm Design and Analysis

Page 2: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 2

Outline

Basic concepts; Case study; Analyzing algorithms Comparing algorithms Problem complexity

Page 3: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 3

Algorithm - Definition

Sequence of steps to carry out a particular task

Examples: Recipe; Assembly instructions; ……

Page 4: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 4

Characteristics of Algorithms

Algorithms are precise. Each step has a clearly defined meaning;

Algorithms are effective. The task is always done as required;

Algorithms have a finite number of steps; Algorithms must terminate.

Page 5: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 5

Any algorithm can be stated using the following logic constructs

Sequence: steps must follow each other in a logical sequence;

Decision: there may be alternative steps that may be taken subject to a particular condition;

Repetition: certain steps may be repeated while, or until, a certain condition is true.

Page 6: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 6

Why study algorithms?

A motivating example…

Page 7: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 7

Example: sorting playing cards

Start with one card on the left hand and

the others cards facing down on the table;

while there are cards left on the table do

{ take one card from the table;

compare it with each of the cards

already in the hand, from left to right,

inserting it into the right position;

}

Sequence

Decision

Repetition

Insertion Sort

Page 8: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 8

Expressing computer algorithms

It should be expressed in a language more precise, less ambiguous, and more compact than a “natural language” such as English;

Algorithms are usually written in a pseudocode language and later translated to a real programming language.

Page 9: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 9

Insertion Sort – An Intuitive Picture

j Q K5

Q

i

1 1

5252

key

A A1

Page 10: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 10

Insertion Sort in Pseudocode

Insertion-Sort(A) ; A is an array of numbers for j = 2 to length(A) key = A[j] i = j - 1 while i > 0 and A[i] > key A[i+1] = A[i] i = i - 1 A[i+1] = key

SequenceRepetition

Decision

Page 11: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 11

Algorithms and Machine Models

Sequential algorithms and machine

models

Parallel algorithms and machine models

Page 12: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 12

Analyzing algorithms

Estimate the running time of algorithms;

= F(Problem Size)

= F(Input Size)

= number of primitive operations used (add, multiply, compare etc)

Page 13: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 13

Analysis for Insertion Sort

Insertion-Sort(A) Cost Times (Iterations) 1 for j = 2 to length(A) { c1

2 key = A[j] c2

3 i = j – 1 c3 4 while i > 0 and A[i] > key c4

5 A[i+1] = A[i] c5

6 i = i – 1 c6

7 A[i+1] = key c7 }

n

n - 1

n - 1

n - 1

Page 14: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 14

Insertion Sort Analysis (cont.)

Best Case:

Array already sorted, tj = 1 for all j

Page 15: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 15

Insertion Sort Analysis (cont.)

Worst Case:

Array in reverse order, tj = j for all j

Note that

Page 16: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 16

Insertion Sort Analysis (cont.)

Average Case:

Check half of array on average,

tj = j/2 for all j

T(n) = a.n2 + bn + c (quadractic in n)

We are usually interested in the worst-case running timeWe are usually interested in the worst-case running time

Page 17: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 18

Growth rates

constant growth rate: T(n) = c linear growth rate : T(n) = c*n logarithmic growth rate : T(n) = c*log n quadratic growth rate : T(n) = c*n2

exponential growth rate : T(n) = c*2n

Page 18: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 19

Asymptotic Analysis

Ignoring constants in T(n) Analyzing T(n) as n "gets large"

The running time grows roughly on the order of

Notationally, =

" n

T n O n

3

3

"

( ) ( )

As grows larger, is MUCH large than , , and ,so it dominates

n n n n nT n

n 23 log( )

T n n n n n n( ) log 13 42 2 43 2Example:

The big-oh (O) Notation

Page 19: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 20

Formally…

T(n) = O(f(n)) if there are constants c and n such that T(n) < c.f(n) when n n0

c.f(n)

T(N)

n0 n

f(n) is an upper bound for T(n)

Page 20: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 21

O ( big-oh)

Describes an upper bound for the running time of an algorithm

Upper bounds for Insertion Sort running times:

•worst case: O(n2) T(n) = c1.n2 + c2.n + c3

•best case: O(n) T(n) = c1.n + c2

•average case: O(n2) T(n) = c1.n2 + c2.n + c3

Time Complexity

Page 21: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 22

Some properties of the O notation

Fastest growing function dominates a sum O(f(n)+g(n)) is O(max{f(n), g(n)})

Product of upper bounds is upper bound for the product If f is O(g) and h is O(r) then fh is O(gr)

f is O(g) is transitive If f is O(g) and g is O(h) then f is O(h)

Hierarchy of functions O(1), O(logn), O(n1/2), O(nlogn), O(n2), O(2n), O(n!)

Page 22: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 23

Times on a 1-billion-steps-per-second computer

Page 23: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 24

Simple statement sequence s1; s2; …. ; sk

O(1) as long as k is constant

Simple loops for(i=0;i<n;i++) { s; } where s is O(1) Time complexity is n O(1) or O(n)

Analyzing an Algorithm

Page 24: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 25

Analyzing an Algorithm (cont.)

Nested loops for(i=0;i<n;i++) for(j=0;j<n;j++) { s; }

Complexity is n O(n) or O(n2)

Recursion Search(n) if middle_element = key return it

else if it’s greater search_left(n/2) else search_right(n/2) Solve recurrence equation: T(n) = T(1) + T(n/2) Complexity is O(log2n)

Page 25: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 26

Reasonable and Unreasonable Algorithms

Polynomial Time algorithms An algorithm is said to be polynomial if it is

O( nc ), c >1 Polynomial algorithms are said to be reasonable

They solve problems in reasonable times!

Exponential Time algorithms An algorithm is said to be exponential if it is

O( rn ), r > 1 Exponential algorithms are said to be unreasonable

Page 26: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 27

Another example: Quick Sort

Uses recursion to repetitively divide list in 1/2 and sort the two halves

QuickSort partitions the list into a left and a right sublist, where all values in the left sublist are less than all values in the right sublist. Then the QuickSort is applied recursively to the sublists until each sublist has only one element.

Page 27: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 28

Quick Sort (cont.)

9 1 25 4 15 4 1 9 25 15

becomes

4 1

25 15 becomes

becomes 1

4 15 25

1 4 15 25

Page 28: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 29

Quick Sort (cont.)

QuickSort (list(0,n)):

If length (list) > 1 then

partition into 2 sublists from

0..split-1 and from split + 1..n

QuickSort (list (0,split-1))

QuickSort (list (split + 1, n)

Page 29: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 30

1. Pick an array value, pivot.2. Use two indices, i and j.

a. Begin with i = left and j =right + 1b. As long as i is less than j do 3 steps:

i. Keep increasing i until we come to an element A[i] >= pivot.ii. Keep decreasing j until we come to an element A[j] <= pivot.iii. Swap A[i] and A[j].

Quick Sort – Partition phase

Page 30: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 31

The Operation of Quicksort

Page 31: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 32

Quick Sort – Complexity

Assume the list size (n) is a power of 2 and that pivot is always chosen such that it divides the list into two equal pieces

At each step we are cutting in half the size to be sorted

O(nlog2n)

Page 32: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 33

Quick Sort demo…

Page 33: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 34

Comparing the Sort Algorithms…

Page 34: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 35

Problem complexity

problems

sort

. . .Insertion Shellsort QuicksortAlgorithms:

Page 35: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 36

Definitions

A problem’s upper bound refers to the best that we know how to do. It is determined by the best algorithmic solution that we have found for a given problem.

Example: For the sorting problem O(n.logn) is the upper bound.

Page 36: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 37

Definitions (cont.)

A problem’s lower bound refers to the best algorithmic solution that is theoretically possible.

Example: It can be proved that sorting an unsorted list cannot be done in less than O(n.log n), unless we make assumptions about the particular input data. Thus O(n.log n) is the lower bound for sorting problems.

Page 37: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 38

Definitions (cont.)

Upper bounds tell us the best we’ve been able to do so far;

Lower bounds tell us the best we can hope to do.

For a given problem, if the upper and lower bounds are the same, then we refer to that problem as a closed problem.

Page 38: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 39

Tractable and Intractable problems

Tractable problems have both upper and lower bounds polynomial – O(logn), O(n), O(n.logn), O(nk);

Intractable problems have both upper and lower bounds exponential - O(2n), O(n!), O(nn);

Page 39: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 40

Problems that cross the line

We have found only exponential solutions (i.e. from the standpoint of our algorithms, it appears to be intractable);

We cannot prove the necessity of an exponential solution (i.e. from the standpoint of our proofs, we cannot say that it is intractable).

Page 40: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 41

NP-complete problems

The upper bound suggests that the problem is intractable;

No proof exists that shows these problems are intractable;

Once a solution is found, it can be checked in polynomial time;

They are all reducible to one another;

Page 41: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 42

Example

Traveling salesman: give a weighted graph (e.g., cities with distances between them) and some distance k (e.g., the maximum distance you want to travel), is there some tour that visits all the points (e.g., cities) and returns home such that distance k?

Page 42: Eleg667/2001-f/Topic-1a 1 A Brief Review of Algorithm Design and Analysis

eleg667/2001-f/Topic-1a 43

P = NP ?

Win the Turing Award

and

get a $1 million prize!http://www.claymath.org/prize_problems/p_vs_np.htm