BNCT Optimization through Parallel Simulated Annealing

I discuss the application of parallel simulated annealing to the problem of finding an optimum tratment plan for Boron Neutron Capture Therapy. Due to the large dimension of the parameter space, and the large cost of evaluating the BNCT cost function it is not feasible to visit every possible treatment to arrive at an optimization plan. Instead a statistical sampling method called simulated annealing is invoked. Simulated annealing is a generalization of the Monte Carlo algorithm that is analogous to a statistical description of physical cooling in an annealing process. The severe computation requirements of simulated annealing involve parallelizing the code across several processors. The code is currently running on a twelve processor Sun computer and a forty processor linux cluster.

Top of Page