What makes machine learning useful for physics? I will provide my answer to this question with examples in many-body physics research. They reveal two general principles. 1) Representation learning. One can use deep neural architectures to extract independent, collective, and useful representations from complex data or microscopic theory. 2) Differentiable programming. One can solve modeling, optimization, control, and inverse design problems in physics by composing differentiable components into a computer program, then tuning the program with gradient optimization.
What can physicists give back to machine learning? There are many. I will focus on a quantum-based probabilistic generative model named “Born Machine”. It holds the promise of delivering the useful advantage of quantum sampling machines.
Professor Lei Wang received his Bachelor degree from Nanjing University in 2006 and Ph.D. from the Institute of Physics, Chinese Academy of Sciences in 2011. He did postdoctoral research on computational quantum physics at ETH Zurich in the next 4.5 years. Prof. Wang joined the Institute of Physics in 2016. His research interest is at the cross-section of machine learning and quantum many-body computation.
Meeting Link: To request for meeting link, please write to firstname.lastname@example.org