Time-lagged autoencoders: Deep learning of slow collective variables for molecular kinetics

TL;DR

This paper introduces a time-lagged autoencoder that effectively reduces the dimensionality of molecular dynamics data, capturing slow collective variables and outperforming linear methods in representing slow stochastic processes.

Contribution

It presents a novel deep learning approach that extends autoencoders with time-lagged features to identify slow dynamics in molecular systems.

Findings

Successfully finds low-dimensional embeddings of molecular data

Captures slow dynamics beyond linear techniques

Reliable in identifying slow collective variables

Abstract

Inspired by the success of deep learning techniques in the physical and chemical sciences, we apply a modification of an autoencoder type deep neural network to the task of dimension reduction of molecular dynamics data. We can show that our time-lagged autoencoder reliably finds low-dimensional embeddings for high-dimensional feature spaces which capture the slow dynamics of the underlying stochastic processes - beyond the capabilities of linear dimension reduction techniques.