A Lightweight Spiking GAN Model for Memristorcentric Silicon Circuit with On-chip Reinforcement Adversarial Learning

As a powerful generative model, Generative Adversarial Network (GAN) is widely studied to automatically generate high-quality new data to greatly enhances the capabilities of artificial intelligence (AI) technology. However, the unique training process of GAN comes at a very high computational complexity and high cost of memory accesses. In this work, a memristor-based spiking-GAN neuromorphic hardware system is proposed to address the challenges. Both the generator and discriminator of GAN are in the form of spiking neural network (SNN) to improve the computational performance, and the memristor synapse circuit with 1 memristor and 4 transistors (1M4T) is proposed as Computing in Memory (CIM) to avoid the cost of memory accesses. The reinforcement learning rule (i.e., reward-modulated spiketiming dependent plasticity, or R-STDP) is used to train both discriminator and generator networks, with a new backpropagation method for the reward/punishment signal. Tests on the MNIST and Fashion-MNIST datasets showed that the proposed GAN can efficiently generate data samples. The results demonstrate the great potential of this memristor-based spiking-GAN for high-speed energy-efficient data augmentations.

Automated summary

This paper proposes a memristor-based spiking-GAN neuromorphic hardware system to address the high computational complexity and memory access cost issues in training Generative Adversarial Network (GAN). By utilizing spiking neural networks (SNN) for the generator and discriminator of GAN, and employing memristor synapse circuits for Computing in Memory (CIM), the system can efficiently generate data samples.