Learning Annealing Schedules for Channel Routing


David C. Hettlinger: Class of 2004

Document Type

Conference Material

Publication Date


Department 1

Computer Science


The choice of a good annealing schedule is necessary for good performance of simulated annealing for channel routing and other combinatorial optimization problems. In this paper, we propose a new means of controlling annealing temperatures by posing the simulated annealing task as an optimal control problem and applying the techniques of reinforcement learning. Although many means of automating control of annealing temperatures have been proposed, this technique requires no specific knowledge of the problem and provides a natural means of expressing time versus quality trade-offs.