Bridging Deep Learning and Many-Body Quantum Physics via Tensor Networks

Explore the intersection of deep learning and many-body quantum physics through tensor networks in this 24-minute APS Physics video. Delve into convolutional and recurrent arithmetic circuits, measures of entanglement for deep learning architectures, and controlling dependencies through layer widths. Examine start-end entanglement in recurrent networks and the exponential memory capacity of deep networks. Discover tensor network constructions of prominent deep learning architectures, and analyze the balance between information re-use and loops. Gain insights into how deep learning architectures support high entanglement, bridging the gap between these complex fields of study.

Intro
Machine Learning and Many-Body Physics
Baseline Architecture - Convolutional Arithmetic Circuit
Baseline Architecture. Convolutional Arithmetic Circuit
Baseline Architecture - Recurrent Arithmetic Circuit
Measures of Entanglement for Deep Learning Archs
Controlling Dependencies -Layer Widths
Start-End Entanglement in Recurrent Networks
Exponential Memory Capacity for Deep Networks
TN Constructions of Prominent Deep Learning Archs
Information Re-Use Vs. Loops
Results - Deep Learning Archs Support High Entanglement