Atomic Nuclei Reveal Limits Of Neural Network Quantum Simulations

Atomic Nuclei Reveal Limits Of Neural Network Quantum Simulations

For a fixed number of configurations, representing quantum states becomes less accurate as their non-stabilizerness increases. This demonstrates a clear limit to how well restricted Boltzmann machines can compress and represent highly entangled systems. Calculations using ground states of medium-mass atomic nuclei reveal non-stabilizerness as a key property governing neural network performance.