The quantum wavefunction presents the ultimate "big data" problem in physics. When many quantum particles interact in a low-temperature material or a quantum computer, the complexity of the quantum state presents a daunting challenge for any classical simulation strategy. Recently, a new computational toolbox based on modern machine learning techniques has been rapidly adopted into the field of condensed matter and quantum information physics. Standard tools like feed-forward and convolutional neural networks are being repurposed for training on "images" of microscopic configurations. Unsupervised and reinforcement learning are making headway in improving standard algorithms such as quantum Monte Carlo. In this talk, I will discuss recent progress, focussing on how generative modelling with stochastic neural networks can be used to combat the complexity of the quantum wavefunction, with applications in materials science, atomic physics, and the design of future quantum computers.
Roger Melko / University of Waterloo
Event Details & Abstracts