SeminarRecording AvailableBiophysics

Neural network-like collective dynamics in molecules

Schedule
Friday, November 27, 2020
14:00 UTC
Arvind Murugan

University of Chicago

Host: Imperial College Physics of Life Network Seminars

Recording

Event Information

Recording

Available

Host

Imperial College Physics of Life Network Seminars

Duration

70 minutes

Abstract

Neural networks can learn and recognize subtle correlations in high dimensional inputs. However, neural networks are simply many-body systems with strong non-linearities and disordered interactions. Hence, many-body physical systems with similar interactions should be able to show neural network-like behavior. Here we show neural network-like behavior in the nucleation dynamics of promiscuously interacting molecules with multiple stable crystalline phases. Using a combination of theory and experiments, we show how the physics of the system dictates relationships between the difficulty of the pattern recognition task solved, time taken and accuracy. This work shows that high dimensional pattern recognition and learning are not special to software algorithms but can be achieved by the collective dynamics of sufficiently disordered molecular systems.

We use essential cookies to run the site. Analytics cookies are optional and help us improve World Wide. Learn more.