分支定界.pdfVIP

2734100190 : xdhu@amss.ac.cn, : : 904 xdhu 1 5. - Exact Methods Divide and Conquer (Searching, Sorting, Traveling Salesman) Dynamic Programming (Shortest path, Polygon Triangulation, Knapsack) Branching and Bound (Integer Linear Programming, Traveling Salesman, Scheduling) Approximation Algorithms or Heuristics Greedy Strategy (Min Spanning Tree, Max Satisfiability, Knapsack, Vertex Cover, Independent Set, Traveling Salesman) Local Search (Max Matching, Traveling Salesman, Max Cut) Sequential Technique (Scheduling, Bin Packing, Vertex Coloring) Integer Programming Approach (Vertex Cover, Max SAT, Max Cut) Randomized Method (Min Cut, Max SAT, Vertex Cover) On-Line Algorithm/Problem (Paging, k-Server, Scheduling, Bin Packing ) xdhu 2 5.3 Branching and Bound Branch-and-Bound is basically a divide-and-conquer method and similar to dynamic programming. It works by maintaining a set S of subinstances Ii of the original instance I of underlying optimization problem such that the union of the feasible solutions of such subinstances is equal to the feasible solutions of I . At the beginning, S is initialized to contain only I . A decomposition operations, called branching, is applied on the problem instance I to derive a set of mutually excluding subinstances Ii . This process is usually modelled as a decision tree whose nodes are associated with subinstances. xdhu 3 5.3 Branching and Bound The method relies on the application of backtracking