Learning efficient erasure protocols for an underdamped memory

TL;DR

This paper uses evolutionary reinforcement learning to discover highly efficient, low-heating erasure protocols for an underdamped mechanical memory, outperforming traditional methods and enabling faster, more reliable memory operations.

Contribution

The study introduces a reinforcement learning approach to optimize erasure protocols for underdamped memories, achieving significant improvements over hand-designed protocols.

Findings

Learned protocols are more efficient than traditional ones.

Protocols enable high-speed, low-heating memory erasure.

Method can be applied to other physics-based systems.

Abstract

We apply evolutionary reinforcement learning to a simulation model in order to identify efficient time-dependent erasure protocols for a physical realization of a one-bit memory by an underdamped mechanical cantilever. We show that these protocols, when applied to the cantilever in the laboratory, are considerably more efficient than our best hand-designed protocols. The learned protocols allow reliable high-speed erasure by minimizing the heating of the memory during the operation. More generally, the combination of methods used here opens the door to the rational design of efficient protocols for a variety of physics applications.