In‐Memory Euclidean Distance Computation in a Stacked Memristor Crossbar for Hardware Self‐Organizing Maps

Jan 14, 2026 by Jinwoo Park, H. Kim (Advanced Functional Materials)

DOI 10.1002/adfm.202531235

We built a 2×32×32 stacked memristor crossbar that computes Euclidean distances in situ by reading the middle electrode current, letting a full SOM pipeline—distance, competition, weight update—run analog and massively parallel without external arithmetic. It’s the first practical demo mapping distance‑driven unsupervised learning directly into a stacked crossbar, which makes SOMs (TSP, image clustering, color quantization) way more scalable and energy efficient than peripheral‑heavy accelerators.

source S2, crossref