Towards Faster Quantum Circuit Simulation Using Graph Decompositions, GNNs and Reinforcement Learning

In this work, we train a graph neural network with reinforcement learning to more efficiently simulate quantum circuits using the ZX-calculus. Our experiments show a marked improvement in simulation efficiency using the trained model over existing methods that do not incorporate AI. In this way, we demonstrate a machine learning model that can reason effectively within a mathematical framework such that it enhances scientific research in the important domain of quantum computing.