PCMNoise: Physical Weight Noise from Phase-Change Memory for Exploration in Deep Reinforcement Learning

IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences 2026

View at publisher
Authors: Zhenhao JIAO, Xiaogang CHEN, Tao HONG, Shunfen Li, Xi LI, Chengcai TU, Weibang Dai, Zhitang SONG
DOI: 10.1587/transfun.2025eal2096
Cited by: 0
References: 12

Abstract

PCMNoise uses standardized phase-change memory conductance readouts as weight noise for exploration in deep RL. The hardware noise is distributionally equivalent to NoisyNet and validated across 10 conductance states. On Atari and MuJoCo, PCMNoise improves final performance and cuts randomness-related training energy by 15.6× versus PRNGs.

Fields of Study & Subjects

Reinforcement Learning in Robotics Advanced Memory and Neural Computing Artificial Intelligence in Games

Publication details

Journal / source
IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences
Volume
E109.A
Issue
8
Pages
1440–1444
Publication date
12 February 2026
Language
EN