Generative thermodynamic computing

TL;DR

This work introduces a generative modeling framework for thermodynamic computing, in which structured data is synthesized from noise by the natural time evolution of a physical system governed by Langevin dynamics, and demonstrates this framework within a digital simulation of a thermodynamic computer.

Abstract

We introduce a generative modeling framework for thermodynamic computing, in which structured data is synthesized from noise by the natural time evolution of a physical system governed by Langevin dynamics. While conventional diffusion models use neural networks to perform denoising, here the information needed to generate structure from noise is encoded by the dynamics of a thermodynamic system. Training proceeds by maximizing the probability with which the computer generates the reverse of a noising trajectory, which ensures that the computer generates data with minimal heat emission. We demonstrate this framework within a digital simulation of a thermodynamic computer. If realized in analog hardware, such a system would function as a generative model that produces structured samples without the need for artificially-injected noise or active control of denoising.

Figures (2)

• Figure 1: (a) Example noising trajectory. (b) The remaining digits used in the training set.
• Figure 2: (a) Three independent dynamical trajectories of the trained denoising thermodynamic computer. (b) The outcome at time $t=t_{\rm f}$ of 25 independent trajectories of the trained computer. (c) Coupling patterns between 16 representative hidden units and the visible layer.