Quantum correlations, formalized in concepts such as entanglement or Bell nonlocality, are at the heart of modern quantum theory, and form the foundation of many of its applications. In certain cases their study becomes analytically difficult and we are forced to use numerical tools to aid us. In recent decades artificial neural networks have proven to have exceptional expressibility and trainability in a wide variety of scenarios. In this talk I aim to display several ways in which neural networks can help us in the study of foundations of quantum theory, with a focus on simple generative models. Such techniques can be helpful when it becomes difficult to work with quantum systems, e.g. when examining many particles or non-convex problems. While showing a few other illustrative use-cases, the primarily demonstrated case study of the use of neural networks will be Bell-nonlocal correlations in networks [1], with the hope of sparking ideas of how such numerics can be useful in assisting theoretical research in other fields.

 

[1]: T. Kriváchy et al, "A neural network oracle for quantum nonlocality problems in networks", npj Quant Inf 6, 70 (2020)