Abstract
Threshold logic has a property that it can represent a complex Boolean function with one threshold logic gate. Recently, many works have focused on the synthesis of threshold logic networks from the Boolean functions. In this paper, we propose a formal approach to the threshold logic network optimization. The proposed approach consists of network decomposition and SAT-based optimization based on a more practical cost function -- the summation of gate counts, weights, and threshold values in the whole threshold network. We conducted the experiments on a set of IWLS 2005 benchmarks. The experimental results show that the proposed algorithm can effectively reduce the cost by nearly 6% for the benchmarks on average.