Stochastic Gradient Adjustment: Single-Pulse In-Memory Training on Phase-Change Memory Chips

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.2025eal2094
Cited by: 0
References: 19

Abstract

This paper introduces a hybrid analog-digital architecture that couples a 40 nm phase-change memory (PCM) crossbar array with a lightweight FPGA-based digital processing unit through a stochastic-gradient adjustment (SGA) rule. On-chip experiments achieve 98.14% accuracy and 0.94 BPC, surpassing PCM+SGD baselines. The SGA framework thus charts a viable path toward large-scale, on-device training for edge intelligence systems.

Fields of Study & Subjects

Phase-change materials and chalcogenides Advanced Memory and Neural Computing Ferroelectric and Negative Capacitance Devices

Publication details

Journal / source
IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences
Volume
E109.A
Issue
6
Pages
1170–1175
Publication date
4 January 2026
Language
EN