Using Recurrent Neural Networks for Circuit Complexity Modeling

Being able to model the complexity of Boolean functions in terms of number of nodes in a binary decision diagram can be quite useful in VLSI/CAD applications. Our investigation showed that it is possible to use the recurrent neural network (RNN) models for the prediction of circuit complexity. The modeling results matched closely with simulations with an average error of less than 1 %. The correlation coefficient between RNN's predictions and actual results for ISCAS benchmark circuits was 0.629.