Evolutionary Neural Networks For Value Ordering In Constraint Satisfaction Problems (1994)
A new method for developing good value-ordering strategies in constraint satisfaction search is presented. Using an evolutionary technique called SANE, in which individual neurons evolve to cooperate and form a neural network, problem-specific knowledge can be discovered that results in better value-ordering decisions than those based on problem-general heuristics. A neural network was evolved in a chronological backtrack search to decide the ordering of cars in a resource-limited assembly line. The network required 1/30 of the backtracks of random ordering and 1/3 of the backtracks of the maximization of future options heuristic. The SANE approach should extend well to other domains where heuristic information is either difficult to discover or problem-specific.
Technical Report AI94-218, Department of Computer Sciences, The University of Texas at Austin.

Risto Miikkulainen Faculty risto [at] cs utexas edu
David E. Moriarty Ph.D. Alumni moriarty [at] alumni utexas net