1 / 65

# algorithms - PowerPoint PPT Presentation

Algorithms. Dynamic Programming. Dynamic Programming. General approach – combine solutions to subproblems to get the solution to an problem Unlike Divide and Conquer in that subproblems are dependent rather than independent bottom-up approach

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.

## PowerPoint Slideshow about 'algorithms' - daniel_millan

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

### Algorithms

Dynamic Programming

• General approach – combine solutions to subproblems to get the solution to an problem

• Unlike Divide and Conquer in that

• subproblems are dependent rather than independent

• bottom-up approach

• save values of subproblems in a table and use more than one time

• Combine “appropriate” subproblem solutions to get a solution to the bigger problem

• If you have a D&C algorithm that does a lot of duplicate computation, you can often find a DP solution that is more efficient

• Calculating binomial coefficients

• n choose k is the number of different combinations of n things taken k at a time

• These are also the coefficients of the binomial expansion (x+y)n

function C(n,k)

if k = 0 or k = n then

return 1

else

return C(n-1, k-1) + C(n-1, k)

C(4,2)

C(4,3)

C(3,1)

C(3,2)

C(3,2)

C(3,3)

• Time complexity is (n!)

• But we did a lot of duplicate computation

• Dynamic programming approach

• Store solutions to sub-problems in a table

• Bottom-up approach

n k

0 1 2 3

0

1

2

3

4

5

n

• Time complexity

O(n k)

• Storage requirements

• full table O(n k)

• better solution

• Dynamic Programming is often used for optimization problems that satisfy the principle of optimality

• Principle of optimality

• In an optimal sequence of decisions or choices, each subsequence must be optimal

• Problem has many possible solutions

• Each solution has a value

• Goal is to find the solution with the optimal value

1 Characterize the structure of an optimal solution

2 Recursively define the value of an optimal solution

3 Compute the value of an optimal solution in a bottom-up manner

4 Construct an optimal solution from computed information

A1A2A3An

• Matrix multiplication is associative

• All ways that a sequence can be parenthesized give the same answer

• But some are much less expensive to compute

• Given a chain < A1,A2, . . .,An> of n matrices, where i = 1, 2, . . ., n and matrix Ai has dimension pi-1pi, fully parenthesize the product A1A2An in a way that minimizes the number of scalar multiplications

Matrix Dimensions

A 13 x 5

B 5 X 89

C 89 X 3

D 3 X 34

M = A B C D 13 x 34

1 ((A B) C) D 10,582

2 (A B) (C D) 54,201

3 (A (B C)) D 2, 856

4 A ((B C) D) 4, 055

5 A (B (C D)) 26,418

1 13 x 5 x 89 to get (A B) 13 x 89 result

13 x 89 x 3 to get ((AB)C) 13 x 3 result

13 x 3 x 34 to get (((AB)C)D) 13 x 34

n 1 2 3 4 5 10 15

T(n) 1 1 2 5 14 4,862 2,674,440

1 Characterize the structure of an optimal solution

2 Recursively define the value of an optimal solution

3 Compute the value of an optimal solution in a bottom-up manner

4 Construct an optimal solution from computed information

• Show that the principle of optimality applies

• An optimal solution to the problem contains within it optimal solutions to sub-problems

• Let Ai..j be the optimal way to parenthesize AiAi+1. . .Aj

• Suppose the optimal solution has the first split at position k

A1..k Ak+1..n

• Each of these sub-problems must be optimally parenthesized

• Define value of the optimal solution recursively in terms of optimal solutions to sub-problems.

• Consider the problem Ai..j

• Let m[i,j] be the minimum number of scalar multiplications needed to compute matrix Ai..j

• The cost of the cheapest way to compute A1..n is m[1,n]

• Define m[i,j]

If i = j the chain has one matrix and the cost m[i,i] = 0 for all i  n

If i < j

Assume the optimal split is at position k where i k < j

m[i,j] = m[i,k] + m[k+1,j] + pi-1 pkpj

to compute Ai..k and Ak+1..j

• Problem: we don’t know what the value for k is, but there are only j-i possibilities

• Compute optimal cost by using a bottom-up approach

• Example: 4 matrix chain

Matrix Dimensions

A1 100 x 1

A2 1 x 100

A3 100 x 1

A4 1 x 100

• po p1 p2 p3 p4

100 1 100 1 100

j

1 2 3 4

1 2 3 4

i

i

1

2

3

4

1

2

3

4

s

m

po p1 p2 p3 p4

100 1 100 1 100

1 n length[p] - 1

2 for i 1 to n

3 do m[i,i] 0

4 for l 2 to n

5 do for i 1 to n - l + 1

6 do ji + l - 1

7 m[i,j] 

8 for k i to j - 1

9 do q m[i,k]+m[k+1,j]+ pi-1 pkpj

10 if q < m[i,j]

11 then m[i,j] = q

12 s[i,j] = k

13 return m and s

• Time complexity

• Triply nested loops

• O(n3)

• Space complexity

• n x n two dimensional table

• O(n2)

• Matrix-Chain-Order determines the optimal number of scalar multiplications

• Does not directly compute the product

• Step 4 of the dynamic programming paradigm is to construct an optimal solution from computed information.

• The s matrix contains the optimal split for every level

1 if j > i

2 then X MATRIX-CHAIN-MULTIPLY(A,s,i,s[i,j])

3 Y MATRIX-CHAIN-MULTIPLY(A,s,s[i,j]+1,j)

4 return MATRIX-MULTIPLY(X,Y)

5 else return Ai

Initial call

MATRIX-CHAIN-MULTIPLY(A,s,1,n) where

A = < A1,A2, . . .,An>

### Algorithms

Dynamic Programming Continued

• When the problem can be reduced to several “overlapping” sub-problems.

• All possible sub-problems are computed

• Computation is done by maintaining a large matrix

• Usually has large space requirement

• Running times are usually at least quadratic

• Variation on dynamic programming

• Idea–memoize the natural but inefficient recursive algorithm

• As each sub-problem is solved, store values in table

• Initialize table with values that indicate if the value has been computed

1 n  length(p) - 1

2 for i 1 to n

3 do for j i to n

4 do m[i,j] 

5 return LOOKUP-CHAIN(p,1,n)

1 if m[i,j] 

2 then return m[i,j]

3 if i = j

4 then m[i,j] 

5 else for k i to j - 1

6 do q LOOKUP-CHAIN(p,i,k) +

LOOKUP-CHAIN(p,k+1,j) +

pi-1pkpj

7 if q < m[i,j]

8 then m[i,j] q

9 return m[i,j]

Space and Time Requirementsof Memoized Version

• Running time (n3)

• Storage (n2)

• Definition 1: Subsequence

Given a sequence

X = < x1, x2, . . . , xm>

then another sequence

Z = < z1, z2, . . . , zk>

is a subsequence of X if there exists a strictly increasing sequence <i1, i2, . . . , ik> of indices of x such that for all j = 1,2,...k we have xij= zj

X = <A,B,D,F,M,Q>

Z = <B, F, M>

Z is a subsequence of X with index sequence <2,4,5>

• Definition 2: Common subsequence

• Given 2 sequences X and Y, we say Z is a common subsequence of X and Y if Z is a subsequence of X and a subsequence of Y

• Definition 3: Longest common subsequence problem

• Given X = < x1, x2, . . . , xm> and Y = < y1, y2, . . . , yn> find a maximum length common subsequence of X and Y

X = <A,B,C,B,D,A,B>

Y = <B,D,C,A,B,A>

1 for every subsequence of X

2 Is there subsequence in Y?

3 If yes, is it longer than the longest subsequence found so far?

Complexity?

• Definition 4: Prefix of a subsequence

If X = < x1, x2, . . . , xm> , the ith prefix of X for i = 0,1,...,m is Xi = < x1, x2, . . . , xi>

• Example

• if X = <A,B,C,D,E,F,H,I,J,L> then

X4 = <A,B,C,D> and X0 = <>

• Theorem 16.1 Optimal Substructure of LCS

Let X = < x1, x2, . . . , xm> and Y = < y1, y2, . . . , yn> be sequences and let Z = < z1, z2, . . . ,zk> be any LCS of X and Y

1. if xm = yn then zk = xm = yn and zk-1 is an LCS of xm-1 and yn-1

2. if xm  yn and zk  x m Z is an LCS of Xm-1 and Y

3. if xm  yn and zk  yn Z is an LCS of Xm and Yn-1

• Case 1

• if xm = yn then there is one sub-problem to solve

find a LCS of Xm-1 and Yn-1 and append xm

• Case 2

• if xm  yn then there are two sub-problems

• find an LCS of Xmand Yn-1

• find an LCS of Xm-1 and Yn

• pick the longer of the two

• Cost is length of the common subsequence

• We want to pick the longest one

• Let c[i,j] be the length of an LCS of the sequences Xi and Yj

• Base case is an empty subsequence--then c[i,j] = 0 because there is no LCS

• Table c[0..m, 0..n] stores the length of an LCS of Xi and Yj c[i,j]

• Table b[1..m, 1..n] stores pointers to optimal sub-problem solutions

C B R F T S Q

0 1 2 3 4 5 6 7

0

A 1

B 2

D 3

F 4

M 5

Q 6

X

Matrix C

j

C B R F T S Q

1 2 3 4 5 6 7

i

A 1

B 2

D 3

F 4

M 5

Q 6

X

Matrix B

1 m length[X]

2 n length[Y]

3 for i  to m

4 do c[i,0] 

5 for j  to n

6 do c[0,j] 

7 for i  to m

8 do for j  to n

9 do if xi = yj

then c[i,j] c[i-1,j-1] + 1

b[i,j] ””

12 else if c[i-1,j] c[i,j-1]

13 then c[i,j] c[i-1,j]

14 b[i,j] ””

15 else c[i,j] c[i,j-1]

16 b[i,j] ””

17 return c and b

1if i = 0 or j = 0

2 then return

3 if b[i,j] = ””

4 then PRINT-LCS(b,X,i-1,j-1)

5 print xi

6 else if b[i,j] ””

7 then PRINT-LCS(b,X,i-1,j)

8 else PRINT-LCS(b,X,i,j-1)

• Time complexity

• Space complexity

• Improved space complexity?

• Turns out to be very similar to matrix chain multiplication

• Mapping one problem to another kind of problem for which you already have a good solution is a useful concept

• A polygon is a piecewise-linear, closed curve in the plane.

• The “pieces” are called sides.

• A point joining two consecutive sides is called a vertex.

• The set of points in the plane enclosed by a simple polygon forms the interior of the polygon.

• the set of point ons the polygon forms its boundary

• the set of points surrounding the polygon form its exterior

• A simple polygon is convex if given any two points on its boundary or in its interior, all points on the line segment drawn between them are contained in the polygon’s boundary or interior.

v5

v0= vn

v4

v1

v3

P = <v0, v1, . . . vn-1>

v2

v5

v0

v5

v0

v4

v4

v1

v1

v3

v3

v2

v2

• A triangulation of a polygon is a set T of chords of the polygon that divide the polygon into disjoint triangles.

• No chords intersect

• The set of chords T is maximal; every chord not in T intersects some chord in T

• Every triangulation of an n-vertex convex polygon

• has n-3 chords

• divides polygon into n-2 triangles

• Given a convex polygonP = <v0, v1, . . . vn-1> and a weight function w defined on triangles formed by sides and chords of P.

• Find a triangulation that minimizes the sum of the weights of the triangles in the triangulation

• One possible weight function is to minimize the sum of the lengths of the sides

• We will show how to view both parenthesization and triangulation as parse trees and show the correspondence between the two views.

• The result will be that a slight modification of the matrix chain problem can be used to solve the triangulation problem.

((A1(A2A3)(A4(A5A6)))

A1(A2A3)

A4(A5A6)

A1

A2A3

A4

A5A6

A2

A3

A5

A6

v0

v6

v1

v5

v2

v4

v3

((A1(A2A3)(A4(A5A6)))

v0

v6

v1

v5

v2

v4

v3

A1 A2 A3 A4 A5 A6

6 matrices

7 dimensions

p0 p1 p2 p3 p4 p5 p6

v0

v6

v1

v5

v2

v4

6 sides

7 vertices

v3