Parallel Simulated Annealing

1
Parallel Simulated Annealing

• Karthik Raman
• Supercomputer Education Research Centre

2
Simulated Annealing

• Heuristic for hard combinatorial optimisation
  problems or continuous optimisation
• Borrowed idea from metallurgy annealing of
  metals
• Perturb configuration of system
• Accept moves that reduce cost with high
  probability
• Accept those that increase cost with low
  probability

3
Simulated Annealing
SA attempts to search for global minimum by
allowing the jump from local minimum
4
Key Annealing Parameters

• Temperature rate of cooling
• Length of run
• Stopping conditions
• Often decided by trial-and-error

5
Why parallelise?

• Large number of iterations (moves)
• SA still requires large amount of resources and
  time
• Probable better examination of search space?
• Can give significant performance benefits (for
  large problem sizes)
• Challenge because SA is essentially sequential!

6
Strategies for parallelisation

• Data-level Processors work on subsets, doing a
  sequential SA not very good
• Maximise search space explored
• Adaptive search neighbourhood refinement
• Genetic crossover between solutions

7
Applications

• TSP and hence other problems that map onto it
• VLSI Placement of circuit components in a cell
• Tertiary Structure of proteins energy
  minimisation

8
References

• Sanvicente-Sánchez H and Frausto-Solís J, MPSA
  A Methodology to Parallelize Simulated Annealing
  and its Application to the Traveling Salesman
  Problem
• Lee F-H A (1995), Parallel Simulated Annealing
  on a Message-Passing Multi-computer, PhD Thesis,
  Utah State University
• Yoshida T, Hiroyasu T, Miki M, Ogura M, and
  Okamoto Y (2002), Energy Minimization of Protein
  Tertiary Structure by Parallel simulated
  Annealing using Genetic Crossover, Genetic and
  Evolutionary Computation Conference (GECCO-2002),
  New York City, New York, USA, GECCO-2002 Workshop
  Program pp.49-52.
• Miki M, Hiroyasu T, Kasai M and Ono K and Jitta T
  (2002), Temperature Parallel Simulated Annealing
  with Adaptive Neighborhood for Continuous
  Optimization Problem Computational Intelligence
  and Applications 149-154
• Mutalik P P, Knight L R, Blanton J L and
  Wainwright R L (1992) Solving Combinatorial
  Optimization Problems Using Parallel Simulated
  Annealing and Parallel Genetic Algorithms