Aging-aware Lifetime Enhancement for Memristor-based Neuromorphic Computing
Shuhang Zhang1,2,d, Grace Li Zhang1,a, Bing Li1,b, Hai (Helen) Li2,3,e and Ulf Schlichtmann1,c
1Chair of Electronic Design Automation
agrace-li.zhang@tum.de
bb.li@tum.de
culf.schlichtmann@tum.de
2Institute for Advanced Study, Technical University of Munich (TUM), Munich, Germany
dshuhang.zhang@tum.de
3Department of Electrical and Computer Engineering, Duke University, Durham, NC, United States
ehai.li@duke.edu
ABSTRACT
Memristor-based crossbars have been applied successfully to accelerate vector-matrix computations in deep neural networks. During the training process of neural networks, the conductances of the memristors in the crossbars must be updated repetitively. However, memristors can only be programmed reliably for a given number of times. Afterwards, the working ranges of the memristors deviate from the fresh state. As a result, the weights of the corresponding neural networks cannot be implemented correctly and the classification accuracy drops significantly. This phenomenon is called aging, and it limits the lifetime of memristor-based crossbars. In this paper, we propose a co-optimization framework combining software training and hardware mapping to reduce the aging effect. Experimental results demonstrate that the proposed framework can extend the lifetime of such crossbars up to 11 times, while the expected accuracy of classification is maintained.
