Abstract
Computing in memory (CIM) technology has shown promising results in reducing the energy consumption of a battery-powered device. On the other hand, to reduce MAC operations, Binary neural networks (BNN) show the potential to catch up with a full-precision model. This paper proposes a robust BNN model applied to the CIM framework, which can tolerate analog noises. These analog noises caused by various variations, such as process variation, can lead to low inference accuracy. We first observe that the traditional batch normalization can cause a BNN model to be susceptible to analog noise. We then propose a new approach to replace the batch normalization while maintaining the advantages. Secondly, in BNN, since noises can be removed when inputs are zeros during the multiplication and accumulation (MAC) operation, we also propose novel methods to increase the number of zeros in a convolution output. We apply our new BNN model in the keyword spotting application. Our results are very exciting.