1 / 20

Algorithms: Design and Analysis Summer School 2013 at VIASM: Random Structures and Algorithms

Algorithms: Design and Analysis Summer School 2013 at VIASM: Random Structures and Algorithms. Lecture 4: Dynamic Programming Phan Thị Hà Dương. Lecture 4: Dynamic Programming. 0. Introduction 1. Matrix-chain multiplication 2* Longest common subsequence 3. All-Pairs shortest paths.

dawsond
Télécharger la présentation

Algorithms: Design and Analysis Summer School 2013 at VIASM: Random Structures and Algorithms

An Image/Link below is provided (as is) to download presentation 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. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Algorithms: Design and AnalysisSummer School 2013 at VIASM: Random Structures and Algorithms Lecture 4: Dynamic Programming Phan Thị Hà Dương

  2. Lecture 4: Dynamic Programming 0. Introduction 1. Matrix-chain multiplication 2* Longest common subsequence 3. All-Pairs shortest paths

  3. 0. Introduction Compare to Divide and Conquer algorithm: - Divide to sub problems (independently) - Conquer sub problems recursively - Combine results of sub problems to resolve initial problem. Problems: If the sub problems are not independent, there are common sub sub problem Then we use a table to stock results of common sub sub problems

  4. Divide and Conquer Algorithm: From up to bottom. See the big problem, divide it and conquer the sub problems • Dynamic Programming: From bottom to up. Beginning with small problem, construct greater problem, up to the original problem.

  5. A very simple example Problem: Calculate C(n,k) = n!/(n-k)!k! • Devide and conquer algorithm: function C(n,k) if (k=0 ou k=n) then return 1; else return C(n-1,k)+C(n-1,k-1); Question: What is the complexity of this algorithm ? Exercise: Write an dynamic programming algorithm to solve this problem. What is its complexity ?

  6. Pascal’s Triangle 0. 1 • 1 1 • 1 2 1 • 1 3 3 1 • 1 4 6 4 1 • 1 5 10 10 5 1 Write an algorithm to the problem by using Pascal’s triangle.

  7. 1. Matrix-chain multiplication Problem: given a chain A1, A2, ..., An of n matrices, where for i = 1, 2, ..., n, matrix Ai has dimension p(i-1) × p(i). Fully parenthesize the product A1* A2*…* An in a way that minimizes the number of scalar multiplications.

  8. Example: M=ABCD, A=(10,2); B=(2,100); C=(100,3); D=(3,20) Number of multiplications (with different way to parenthesize)

  9. Analyzing the Problem Example: M=ABCD, A=(10,2); B=(2,100); C=(100,3); D=(3,20) Number of multiplications (with different way to parenthesize) ((AB)C)D: 10x2x100 +2x100x3 +10x3x20 =3200 muls. (AB)(CD): 10x2x100 +100x3x20 +10x100x20 = 28000 (A(BC))D: 2x100x3 +10x2x3 +10x3x20 = 1260 A((BC)D): 2x100x3 +2x3x20 +10x2x20 = 1120 A(B(CD)): 100x3x20 +2x100x20 +10x2x20 = 10400

  10. MATRIX-MULTIPLY(A, B) MATRIX-MULTIPLY(A, B) 1 if columns[A] <> rows[B] 2 then error "incompatible dimensions" 3 else for i ← 1 to rows[A] 4 do for j ← 1 to columns[B] 5 do C[i, j] ← 0 6 for k ← 1 to columns[A] 7 do C[i, j] ← C[i, j] + A[i, k] * B[k, j] 8 return C

  11. The number of ways to parenthesize • If there n matrix, this number is the number of Catalan C(n) = 1/n x C(2n-2,n-1) = Ω (4^n/n^2). • We can not consider all the case to make decision. • Idea: If for computing M, a cut at position i is optimal, then an minimum cost is composed of this cut and minimum cost for computing Ai … Aj and minimum cost for computing A(i+1) … An • Use a table m(i,j) to stock the minimum cost for computing Ai … Aj.

  12. Computing the minimum cost Computing m(i,j): if we cut at k, then m[i,j]=m[i,k]+m[k+1,j]+p(i-1) p(k) p(j). So the formula for m(i,j) is We compute m(i,j) by diagonal: l = j-i+1, the length of the chain of matrix. We compute m(I,j) for l from 2 to n.

  13. Example M=ABCD, A=(10,2); B=(2,100); C=(100,3); D=(3,20). L=2:a[12]=2000,a[23]=600,a[34]=18000 … j = 1 2 3 4 i =1 0 2000 ? ? 2 0 600 ? 3 0 18000 4 0

  14. Algorithm MATRIX-CHAIN-ORDER(p) 1 n ← length[p] 2 for i ← 1 to n 3 do m[i, i] ← 0 4 for l ← 2 to n ▹ l is the chain length. 5 do for i ← 1 to n - l + 1 6-7 do j ← i + l – 1 ; m[i, j] ← ∞ 8 for k ← i to j - 1 9 do q←m[i,k]+m[k+1,j]+p(i-1) p(k) p(j) 10 if q < m[i, j] 11-12 then m[i, j] ← q; s[i, j] ← k 13 return m and s

  15. 3. All-Pairs shortest paths Problem: Given a graph G= (V, E) where each edge having a length. Find the shortest path for all pairs of V.

  16. Floyd-Warshall algorithm Notation: V={1, 2, .., n} Length of edges: l[i,i] = 0, l[i,j] = l(e) if (i,j) in E l[i,j] = ∞ otherwise Distance: d(i,j) is the temporal shortest length from i to j.

  17. Idea of FW’s algorithm • If k is on a shortest path form I to j, then the sub paths from I to k and from k to j are shortest. • Algorithm • Beginning: d = l • After step k, d(i,j) is the shortest path from i to j (path containing only vertices 1, 2, …, k) • After n steps, d(i,j) is the shortest path from i to j

  18. Example 0 5 ∞ ∞ D0=L= 50 0 15 5 30 ∞ 0 15 15 ∞ 5 0 0 5 ∞ ∞ D1= 50 0 15 5 30 35 0 15 15 20 5 0 0 5 20 10 0 5 20 10 0 5 15 10 D2= 50 0 15 5 D3= 50 0 15 5 D4= 20 0 10 5 30 35 0 15 30 35 0 15 30 35 0 15 15 20 5 0 15 20 5 0 15 20 5 0 4 1 15 5 5 50 15 5 30 2 3 15

  19. FW’s algorithm Floyd(L) 1 array D = L 2 for (k=1 to n) 3 dofor (i=1 to n) 4 do for (j=1 to n) 5 do D[i,j]= min(D[i,j],D[i,k]+D[k,j]); 6 return D;

  20. Exercise • What is the complexity of the FW’s algorithm ? • Prove the correctness of the algorithm. • Write an algorithm which return not only the length of shortest path but also a shortest path for each pair of vertices. • Write an algorithm to determine if there exist a path from each pair of vertices.

More Related