Even Distribution Evaluation in Random Stimulus Generation

Abstract

This paper has two contributions: First is to analyze the entropy evaluation for random stimulus generation in one paper of DATE 2008 [1]; second is to present better methods to evaluate the solutions' even distribution for random stimulus generation. An evaluation strategy called min-distance-sum takes the solution space as a ring and calculates the adjacent solutions' distances. Experimental results for SAT and circuit benchmarks showed that our method can evaluate the even degree better than entropy, and also prove that the XOR-based method can improve the distribution of solutions for random stimulus generation.