Dynamic Programming 动态规划 最优化理论与方法课件（英文版－南京大学）.ppt

Dynamic Programming动态规划 * Two Well-known Puzzles Example 1 Suppose there are 30 matches on a table. I begin by picking up 1, 2 or 3 matches. Then my opponent must pick up 1, 2 or 3 matches. We continue in this fashion until the last match is picked up. The player who picks up the last match is the loser. How can I (the first player) be sure of winning the game? Example 2 I have a 9-liter and a 4-liter cup. If I want to bring home exactly 6 liters of milk, how can I accomplish this? * A Network Problem Joe lives in Newyork city and plans to drive to Los Angeles. What is the shortest route that he can travel? Many applications of DP reduce to finding the shortest path (or longest) that joins two points in a given network. * A Network Problem Let’s find the shortest path between Newyork and LA by going backward. * A Network Problem * A Network Problem – Stage 4 We first determine the shortest path to Los Angeles from each stage 4 city. * A Network Problem – Stage 3 We now go backward one stage and find the shortest path from each stage 3 city. * A Network Problem – Stage 3 * A Network Problem – Stage 2 Path 3: * A Network Problem – Stage 1 Optimal Path: 1?2?5?8?10 * Computational Efficiency of DP For the above network example, it is easy to enumerate all the possible paths. (There are 3x3x2=18 paths). For larger problems, DP is much more efficient than the enumeration of all paths. Suppose, we have network consisting of 7 stages, and stage 2 through stage 6 consisting of 5 nodes each. * Computational Efficiency of DP There are 55 possible paths from node 1 to node 27. It requires 5 additions to determine the length of each path: Therefore, the total number of additions needed in explicit enumeration is 5(55)=56 Suppose we use DP to determine the shortest path from node 1 to node 27, how many additions are required? * Computational Efficiency of DP In total, DP requires 4(25)+5=105 additions to find the shortest path from node 1 to node 27 Beside