Characterizing quantum computers with quantum-physics-aware neural networks

Learning the errors in many-qubit quantum computing systems is difficult because of the large parameter space of possible errors, the complex crosstalk errors that occur in many systems, and the ubiquity of non-Markovian effects that are beyond the standard, process matrix model used in most error characterization methods. In this talk, I will present quantum-physics-aware neural networks (QPANNs), which are a kind of neural network that is designed for scalable characterization of complex errors in quantum computers. QPANNs contain interpretable parameters corresponding to sparse Lindblad coefficients that, after the network is trained, explain the errors that are occurring and their rates. We demonstrate these networks’ ability to learn coherent crosstalk errors and context-dependent errors in simulations, and demonstrate their application to experimental systems. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525