Fast Probabilistic Modeling for Combinatorial Optimization

Probabilistic models have recently been utilized for the optimization of large combinatorial search problems. However, complex probabilistic models that attempt to capture interparameter dependencies can have prohibitive computational costs. The algorithm presented in this paper, termed COMIT, provides a method for using probabilistic models in conjunction with fast search techniques. We show how COMIT can be used with two very different fast search algorithms: hillclimbing and Population-based incremental learning (PBIL). The resulting algorithms maintain many of the benefits of probabilistic modeling, with far less computational expense. Extensive empirical results are provided; COMIT has been successfully applied to jobshop scheduling, traveling salesman, and knapsack problems. This paper also presents a review of probabilistic modeling for combinatorial optimization. 1 Background Within the past few years, there have been several novel methods proposed for probabilistic modeling f...

• Type: Conference
• Year: 1998
• Where: AAAI
• Authors: Shumeet Baluja, Scott Davies
• Keywords: AAAI 1998, Intelligent Agents, Probabilistic, Probabilistic Modeling, Probabilistic Models