Evolutionary Computing and the Traveling Salesman By Sam Mulder
Overview Local optimization for the TSP Evolutionary Computing for the TSP Combining local search with EAs Proposed algorithm Neural algorithm for comparison Results Conclusions
Local Optimization for the TSP Lin-Kernighan Algorithm Double bridges Chained Lin-Kernighan Algorithm
Double Bridge
Evolutionary Computing for the TSP Non-Lamarkian EAs Mutation and Crossover Very small problem sizes Low quality results
Combining Local Search with EAs Chained Lin-Kernighan revisited Visualization of search space Adding Lamarkian learning Scatter Search
Proposed Algorithm Population Crossover Mutation Learning Competition
Neural Algorithm for Comparison Divide and Conquer using ART Self-Organizing Maps on sub-problems Lin-Kernighan optimization Merging tours
Results Speed vs Quality 100 city problem 1000 city problem
Conclusions EA = parallel Chained Lin-Kernighan Parallel implementation may allow scaling Local Search + EA = unsolved problem Need to try divide and conquer + EA
Results cont. Chained Lin-Kernighan Proposed Algorithm Neural Clustering