Machine learning the thermodynamic arrow of time

TL;DR

This paper demonstrates that machine learning can identify the thermodynamic arrow of time from microscopic trajectories, matching theoretical bounds and uncovering underlying physical mechanisms.

Contribution

It introduces a machine learning approach that learns to determine the direction of time's arrow and reveals the physical observables involved, advancing understanding of nonequilibrium systems.

Findings

Algorithm's performance matches theoretical bounds

Reveals physical observables related to time's arrow

Discovers thermodynamic mechanisms from data

Abstract

The mechanism by which thermodynamics sets the direction of time's arrow has long fascinated scientists. Here, we show that a machine learning algorithm can learn to discern the direction of time's arrow when provided with a system's microscopic trajectory as input. The performance of our algorithm matches fundamental bounds predicted by nonequilibrium statistical mechanics. Examination of the algorithm's decision-making process reveals that it discovers the underlying thermodynamic mechanism and the relevant physical observables. Our results indicate that machine learning techniques can be used to study systems out of equilibrium, and ultimately to uncover physical principles.