The Coordination of Parallel Search with Common Components

1
The Coordination of Parallel Search with Common
Components

• Stephen Chen and Gregory PittYork University

2
Heuristic Search

• Search Operators
• Control Strategies

3
An Obvious Failing

• Point search operators do not have historical
  information!

4
Search Operator Strategies

• What can be improved?
• What is worth keeping?

5
Random Local Minima
6
Simulated Annealing
7
Summary

• Random search space, random search strategy
• All the intelligence of simulated annealing is in
  the control strategy

8
Distribution of Local Minima (Boese)
Length
Average distance from other local minima
9
Big-Valley Globally Convex
10
Directed Simulated Annealing
11
Coordination with Common Components
12
Restricted Mutations

• Parent 1 1 1 1 0 0 1 0 1 1 1
• Parent 2 1 0 1 1 0 1 1 0 1 1
• Common 1 1 0 1 1 1
• Uncommon 1 1 0 0 1
• Uncommon 2 0 1 1 0

13
TSP Results
Instance n1 n2 n4 n8
pcb1173 8.55 8.07 5.08 8.05 4.88 7.49 4.40
fl1400 5.82 4.33 3.15 3.58 2.45 2.86 1.94
fl1577 10.97 9.09 2.48 7.69 2.19 6.91 1.49
14
Summary

• Globally convex search space, respectful search
  operator
• Add some intelligence to the search operator

15
Conclusion

• The preservation of common components can improve
  the performance of parallel search procedures in
  globally convex search spaces