Fault-tolerant training of neural networks in the presence of MOS transistor mismatches

Analog techniques are desirable for hardware implementation of neural networks due to their numerous advantages such as small size, low power, and high speed. However, these advantages are often offset by the difficulty in the training of analog neural network circuitry. In particular, training of the circuitry by software based on hardware models is impaired by statistical variations in the integrated circuit production process, resulting in performance degradation. In this paper, a new paradigm of noise injection during training for the reduction of this degradation is presented. The variations at the outputs of analog neural network circuitry are modeled based on the transistor-level mismatches occurring between identically designed transistors. Those variations are used as additive noise during training to increase the fault tolerance of the trained neural network. The results of this paradigm are confirmed via numerical experiments and physical measurements and are shown to be superior to the case of adding random noise during training.