Training thermodynamic computers by gradient descent
This work enables the application of machine learning techniques to thermodynamic computing, potentially reducing energy costs for specific computations, though it appears incremental as it adapts existing gradient descent methods to a new hardware paradigm.
The authors tackled the problem of training thermodynamic computers by proposing a gradient descent method that adjusts parameters to perform computations at a specified time, using a teacher-student scheme to mimic neural network activations. They demonstrated this on an image-classification task, achieving a thermodynamic advantage exceeding seven orders of magnitude in energy efficiency.
We show how to adjust the parameters of a thermodynamic computer by gradient descent in order to perform a desired computation at a specified observation time. Within a digital simulation of a thermodynamic computer, training proceeds by maximizing the probability with which the computer would generate an idealized dynamical trajectory. The idealized trajectory is designed to reproduce the activations of a neural network trained to perform the desired computation. This teacher-student scheme results in a thermodynamic computer whose finite-time dynamics enacts a computation analogous to that of the neural network. The parameters identified in this way can be implemented in the hardware realization of the thermodynamic computer, which will perform the desired computation automatically, driven by thermal noise. We demonstrate the method on a standard image-classification task, and estimate the thermodynamic advantage -- the ratio of energy costs of the digital and thermodynamic implementations -- to exceed seven orders of magnitude. Our results establish gradient descent as a viable training method for thermodynamic computing, enabling application of the core methodology of machine learning to this emerging field.