Learning Evaluation Functions for Global Optimization and Boolean Satisfiability

This paper describes STAGE, a learning approach to automatically improving search performance on optimization problems.STAGElearns an evaluation function which predicts the outcomeof a local search algorithm, such as hillclimbing or WALKSAT, as a function of state features along its search trajectories. Thelearned evaluation function is used to bias future search trajectories toward better optima. Wepresent positive results on six large-scale optimizationdomains.

• Type: Conference
• Year: 1998
• Where: AAAI
• Authors: Justin A. Boyan, Andrew W. Moore
• Keywords: AAAI 1998, Evaluation Function, Future Search Trajectories, Intelligent Agents, Search Trajectories