Reinforcement Learning for Scalable Logic Optimization with Graph Neural Networks
This addresses the problem of scalable logic optimization for circuit design, representing an incremental improvement over existing methods.
The paper tackles the NP-hard problem of logic optimization by combining graph convolutional networks with reinforcement learning and a novel scalable node embedding method to learn local transforms for logic graphs, achieving similar size reduction as ABC on smaller circuits and outperforming it by 1.5-1.75x on larger random graphs.
Logic optimization is an NP-hard problem commonly approached through hand-engineered heuristics. We propose to combine graph convolutional networks with reinforcement learning and a novel, scalable node embedding method to learn which local transforms should be applied to the logic graph. We show that this method achieves a similar size reduction as ABC on smaller circuits and outperforms it by 1.5-1.75x on larger random graphs.