Adjacent swap mutation modifies a TSP tour by exchanging vertices that are next to each other in its representation. It is a small-range local-search operator used within mutation. section 18.2.4
Its practical trade-offs are:
- It makes only a local change, so escaping local minima is harder than with mutation that swaps random vertex pairs. section 18.2.4
- Experiments vary adjacent-swap probability alongside crossover and random-swap probability, measuring the resulting average tour cost. section 18.2.4
TSP mutation comparison
Mutation using adjacent swaps eventually plateaus more than the more effective random-pair mutation in the reported TSP experiments. section 18.2.4
Figure 18.16 How TSP population evolves when only mutations versus only crossover are enabled. Crossover-only evolution plateaus soon enough, while the mutation operator swapping random pairs is the most successful. On the x axis, there is the number of generations, while on the y axis is the cost of the tour.
Compare how mutation-only and crossover-only TSP populations evolve across generations and tour cost. section 18.2.4
