[21a-C201-1~11] FS.1 Focused Session "AI Electronics"

[21a-C201-2] Investigation on Non-Ideality of Non-Volatile Memory Array Elements in Neural Network Inference with Computation-in-Memory

〇Kazuhide Higuchi¹, Chihiro Matsui¹, Ken Takeuchi¹ (1.Univ. of Tokyo)

Keywords:Computation-in-Memory, Non Volatile Memory, Neural Network

Computation-in-Memory (CiM) elements have non-idealities, which may degrade the inference accuracy in neural networks. Therefore, we propose a simulation model to handle not only the input data and the network model, but also the memory array non-idealities that cause inference degradation.