Multilevel logic optimization using genetic programming technique

We apply the genetic programming to the multilevel logic optimization problem whose object is acquisition of logic expressions that satisfy a truth table. In our method, a LISP S-expression is used as an individual, and a population consists of individuals for each output. The fitness function reflects the accuracy and the size of an S-expression, and populations for all outputs are repeatedly evolved to minimize fitness values. In order to obtain S-expressions with small size, we employ five kinds of factorization and common sub-expression identification. The proposed method has been implemented and some experimental results have been shown.