Learning Quantum with Generative Models

Explore the intersection of quantum physics and machine learning in this 49-minute talk by Juan Carrasquilla from the Vector Institute for Artificial Intelligence. Delve into the use of generative models for learning quantum states, covering topics such as machine learning in many-body physics, fluctuations in handwritten digits (MNIST), and the extension of restricted Boltzmann machines to the quantum domain. Discover neural network quantum states and the need to go beyond standard quantum state tomography. Examine results on synthetic datasets for GHZ states and investigate recurrent neural network models. Gain insights into the numerical investigation of sample complexity when learning ground states of local Hamiltonians.

Intro
MACHINE LEARNING/MANY-BODY PHYSICS FRIENDS
FLUCTUATIONS HANDWRITTEN DIGITS (MNIST)
ANALYTICAL UNDERSTANDING WHAT DOES THE CAN USE TO MAKE PREDICTIONS?
EXTENDING RESTRICTED BOLTZMANN MACHINES TO THE QUANTUM DOMAIN
NEURAL NETWORK QUANTUM STATES
NEED TO GO BEYOND STANDARD QUANTUM STATE TOMOGRAPHY
RESULTS ON SYNTHETIC DATASETS FOR GHZ STATES
RECURRENT NEURAL NETWORK MODEL AND RESULTS
NUMERICAL INVESTIGATION OF THE SAMPLE COMPLEXITY OF LEARNING
GROUND STATES OF LOCAL HAMILTONIANS