Weight Sharing for Array-Efficient CNN Inference in Compute-in-Memory Architectures
Citations

SCOPUS

0

초록

In this work, we propose a weight sharing method for compute-in-memory (CIM)-based convolutional neural network (CNN) inference, which reduces the number of mapped weights without relying on pruning. By sharing weights and layer-wise scale factors, our approach reduces overall CIM array usage while maintaining model accuracy. Evaluations on ResNet-20 with 32×32 arrays demonstrate that our method achieves a 28% reduction in array usage with less than 1% accuracy loss.

키워드

compute-in-memory (CIM)weight mapping
제목
Weight Sharing for Array-Efficient CNN Inference in Compute-in-Memory Architectures
저자
Rhe, JohnnyKo, Jong Hwan
DOI
10.1109/ISOCC66390.2025.11329740
발행일
2025
유형
Conference Paper
저널명
International SoC Design Conference 2025, ISOCC 2025 - Proceedings of Technical Papers