1. DYNAMIC PROGRAMMING

2. 2 Algorithmic Paradigms Greedy. Build up a solution incrementally, myopically optimizing some local criterion. Divide-and-conquer. Break up a problem into two sub-problems, solve each sub-problem independently, and combine solution to sub-problems to form solution to original problem. Dynamic programming. Break up a problem into a series of overlapping sub-problems, and build up solutions to larger and larger sub-problems.

3. 3 Dynamic Programming History Bellman. Pioneered the systematic study of dynamic programming in the 1950s. Etymology. n Dynamic programming = planning over time. n Secretary of Defense was hostile to mathematical research. n Bellman sought an impressive name to avoid confrontation. – "it's impossible to use dynamic in a pejorative sense" – "something not even a Congressman could object to" Reference: Bellman, R. E. Eye of the Hurricane, An Autobiography.

4. 4 Dynamic Programming Applications Areas. n Bioinformatics. n Control theory. n Information theory. n Operations research. n Computer science: theory, graphics, AI, systems, …. Some famous dynamic programming algorithms. n Viterbi for hidden Markov models. n Unix diff for comparing two files. n Smith-Waterman for sequence alignment. n Bellman-Ford for shortest path routing in networks. n Cocke-Kasami-Younger for parsing context free grammars.

5. Sequence Alignment

6. 6 String Similarity How similar are two strings? n ocurrance n occurrence ocurrance ccurrenceo - ocurrnce ccurrnceo --a e- ocurrance ccurrenceo - 5 mismatches, 1 gap 1 mismatch, 1 gap 0 mismatches, 3 gaps

7. 7 Applications. n Basis for Unix diff. n Speech recognition. n Computational biology. Edit distance. [Levenshtein 1966, Needleman-Wunsch 1970] n Gap penalty  ; mismatch penalty  pq. n Cost = sum of gap and mismatch penalties. 2  +  CA CGACCTACCT CTGACTACAT TGACCTACCT CTGACTACAT - T C C C  TC +  GT +  AG + 2  CA - Edit Distance

8. 8 Goal: Given two strings X = x 1 x 2... x m and Y = y 1 y 2... y n find alignment of minimum cost. Def. An alignment M is a set of ordered pairs x i -y j such that each item occurs in at most one pair and no crossings. Def. The pair x i -y j and x i' -y j' cross if i j'. Ex: CTACCG vs. TACATG. Sol: M = x 2 -y 1, x 3 -y 2, x 4 -y 3, x 5 -y 4, x 6 -y 6. Sequence Alignment CTACC- TACAT- G G y1y1 y2y2 y3y3 y4y4 y5y5 y6y6 x2x2 x3x3 x4x4 x5x5 x1x1 x6x6

9. 9 Sequence Alignment: Problem Structure Def. OPT(i, j) = min cost of aligning strings x 1 x 2... x i and y 1 y 2... y j. n Case 1: OPT matches x i -y j. – pay mismatch for x i -y j + min cost of aligning two strings x 1 x 2... x i-1 and y 1 y 2... y j-1 n Case 2a: OPT leaves x i unmatched. – pay gap for x i and min cost of aligning x 1 x 2... x i-1 and y 1 y 2... y j n Case 2b: OPT leaves y j unmatched. – pay gap for y j and min cost of aligning x 1 x 2... x i and y 1 y 2... y j-1

10. 10 Sequence Alignment: Algorithm Analysis.  (mn) time and space. English words or sentences: m, n  10. Computational biology: m = n = 100,000. 10 billions ops OK, but 10GB array? Sequence-Alignment(m, n, x 1 x 2...x m, y 1 y 2...y n, ,  ) { for i = 0 to m M[0, i] = i  for j = 0 to n M[j, 0] = j  for i = 1 to m for j = 1 to n M[i, j] = min(  [x i, y j ] + M[i-1, j-1],  + M[i-1, j],  + M[i, j-1]) return M[m, n] }

11. Shortest Paths

12. 12 Shortest Paths Shortest path problem. Given a directed graph G = (V, E), with edge weights c vw, find shortest path from node s to node t. Ex. Nodes represent agents in a financial setting and c vw is cost of transaction in which we buy from agent v and sell immediately to w. s 3 t 2 6 7 4 5 10 18 -16 9 6 15 -8 30 20 44 16 11 6 19 6 allow negative weights

13. 13 Shortest Paths: Failed Attempts Dijkstra. Can fail if negative edge costs. Re-weighting. Adding a constant to every edge weight can fail. u t sv 2 1 3 -6 st 2 3 2 -3 3 5 5 6 6 0

14. 14 Shortest Paths: Negative Cost Cycles Negative cost cycle. Observation. If some path from s to t contains a negative cost cycle, there does not exist a shortest s-t path; otherwise, there exists one that is simple. st W c(W) < 0 -6 7 -4

15. 15 Shortest Paths: Dynamic Programming Def. OPT(i, v) = length of shortest v-t path P using at most i edges. n Case 1: P uses at most i-1 edges. – OPT(i, v) = OPT(i-1, v) n Case 2: P uses exactly i edges. – if (v, w) is first edge, then OPT uses (v, w), and then selects best w-t path using at most i-1 edges Remark. By previous observation, if no negative cycles, then OPT(n-1, v) = length of shortest v-t path.

16. 16 Shortest Paths: Implementation Analysis.  (mn) time,  (n 2 ) space. Finding the shortest paths. Maintain a "successor" for each table entry. Shortest-Path(G, t) { foreach node v  V M[0, v]   M[0, t]  0 for i = 1 to n-1 foreach node v  V M[i, v]  M[i-1, v] foreach edge (v, w)  E M[i, v]  min { M[i, v], M[i-1, w] + c vw } }

17. 17 Shortest Paths: Practical Improvements Practical improvements. n Maintain only one array M[v] = shortest v-t path that we have found so far. n No need to check edges of the form (v, w) unless M[w] changed in previous iteration. Theorem. Throughout the algorithm, M[v] is length of some v-t path, and after i rounds of updates, the value M[v] is no larger than the length of shortest v-t path using  i edges. Overall impact. n Memory: O(m + n). n Running time: O(mn) worst case, but substantially faster in practice.

18. 18 Bellman-Ford: Efficient Implementation Push-Based-Shortest-Path(G, s, t) { foreach node v  V { M[v]   successor[v]   } M[t] = 0 for i = 1 to n-1 { foreach node w  V { if (M[w] has been updated in previous iteration) { foreach node v such that (v, w)  E { if (M[v] > M[w] + c vw ) { M[v]  M[w] + c vw successor[v]  w } If no M[w] value changed in iteration i, stop. }

19. Segmented Least Squares

20. 20 Segmented Least Squares Least squares. n Foundational problem in statistic and numerical analysis. n Given n points in the plane: (x 1, y 1 ), (x 2, y 2 ),..., (x n, y n ). n Find a line y = ax + b that minimizes the sum of the squared error: Solution. Calculus  min error is achieved when x y

21. 21 Segmented Least Squares Segmented least squares. n Points lie roughly on a sequence of several line segments. n Given n points in the plane (x 1, y 1 ), (x 2, y 2 ),..., (x n, y n ) with n x 1 < x 2 <... < x n, find a sequence of lines that minimizes f(x). Q. What's a reasonable choice for f(x) to balance accuracy and parsimony? x y goodness of fit number of lines

22. 22 Segmented Least Squares Segmented least squares. n Points lie roughly on a sequence of several line segments. n Given n points in the plane (x 1, y 1 ), (x 2, y 2 ),..., (x n, y n ) with n x 1 < x 2 <... < x n, find a sequence of lines that minimizes: – the sum of the sums of the squared errors E in each segment – the number of lines L n Tradeoff function: E + c L, for some constant c > 0. x y

23. 23 Dynamic Programming: Multiway Choice Notation. n OPT(j) = minimum cost for points p 1, p i+1,..., p j. n e(i, j) = minimum sum of squares for points p i, p i+1,..., p j. To compute OPT(j): n Last segment uses points p i, p i+1,..., p j for some i. n Cost = e(i, j) + c + OPT(i-1).

24. 24 Segmented Least Squares: Algorithm Running time. O(n 3 ). n Bottleneck = computing e(i, j) for O(n 2 ) pairs, O(n) per pair using previous formula. INPUT: n, p 1,…,p N, c Segmented-Least-Squares() { M[0] = 0 for j = 1 to n for i = 1 to j compute the least square error e ij for the segment p i,…, p j for j = 1 to n M[j] = min 1  i  j (e ij + c + M[i-1]) return M[n] } can be improved to O(n 2 ) by pre-computing various statistics