Performance Impacts of Analog ReRAM Non-ideality on Neuromorphic Computing

• Published in: IEEE transactions on electron devices Vol. 66; no. 3; pp. 1289 - 1295
• Main Authors: Lin, Yu-Hsuan, Wang, Chao-Hung, Lee, Ming-Hsiu, Lee, Dai-Ying, Lin, Yu-Yu, Lee, Feng-Min, Lung, Hsiang-Lan, Wang, Keh-Chung, Tseng, Tseung-Yuen, Lu, Chih-Yuan
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.03.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Analog memory; Artificial neural networks; Computation; Computer simulation; Error reduction; Inference; Mapping; Memory management; Multilayers; Neural networks; Neuromorphic computing; Neuromorphics; noise; Random access memory; reliability; Resistance; resistive random access memory (ReRAM); stability; Synapses; Training; Variation; Weight
• ISSN: 0018-9383; 1557-9646
• DOI: 10.1109/TED.2019.2894273

Resistive random access memory (ReRAM) is often considered as a strong candidate for storing the weights in non-von Neumann neuromorphic computing systems. This paper studies how nonideal memory characteristics, which include programing error, read fluctuation, and retention, affect the inference accuracy of the analog ReRAM neural networks by incorporating memory characteristics extracted from 1-Mb ReRAM into a simulated inference-only neural network. This paper also shows that the different layer in the network can tolerate different amount of such imperfects. We learned four key points: 1) the conductance range of memory with less relative fluctuation is preferred for designing the weight-conductance mapping; 2) the control of programing error is essential for high inference accuracy; 3) retention-induced conductance drift can be fatal to the neuromorphic system. A compensation scheme is proposed in this paper which can effectively recover the inference accuracy; and 4) for multilayer networks, avoiding weight errors in the front layers can help to maintain the inference accuracy by reducing calculation error which may otherwise accumulate and pass down the networks. The concepts and approaches of this paper can also be applied to evaluate other types of nonvolatile memories for artificial neural networks.