In Situ Quantization with Memory-Transistor Transfer Unit Based on Electrochemical Random-Access Memory for Edge Applications.

Jan 21, 2026 by Zhen Yang, Yuxiang Yang, Baiqian Wang, Yaoyu Tao, Zelun Pan, Lei Cai, Teng Zhang, Longhao Yan, Xianbin Li, Yuch-Chi Yang (Advancement of science)

DOI 10.1002/advs.202521815

We built a compact memory-transistor synaptic unit by integrating ECRAM-style ionic nonvolatile cells with a transfer transistor to perform in situ approximate quantization from the device physics itself, so networks can do low-precision quantization and parallel MVM without extra binary programming. It’s a practical edge-ready scheme that preserves accuracy for binary/ternary nets, supports continual learning with less forgetting, and shows huge energy wins over digital platforms at array scale.

source S2, crossref