Generative algorithms such as GANs are at the cusp of next revolution in the field of unsupervised learning and large-scale artificial data generation. However, the adversarial (competitive) co-training of the discriminative and generative networks in GAN makes them computationally intensive and hinders their deployment on the resource-constrained IoT edge devices. Moreover, the frequent data transfer between the discriminative and generative networks during training significantly degrades the efficacy of the von-Neumann GAN accelerators such as those based on GPU and FPGA. Therefore, there is an urgent need for development of ultra-compact and energy-efficient hardware accelerators for GANs. To this end, in this work, we propose to exploit the passive RRAM crossbar arrays for performing key operations of a fully-connected GAN: (a) true random noise generation for the generator network, (b) vector-by-matrix-multiplication with unprecedented energy-efficiency during the forward pass and backward propagation and (C) in-situ adversarial training using a hardware friendly Manhattan's rule. Our extensive analysis utilizing an experimentally calibrated phenomological model for passive RRAM crossbar array reveals an unforeseen trade-off between the accuracy and the energy dissipated while training the GAN network with different noise inputs to the generator. Furthermore, our results indicate that the spatial and temporal variations and true random noise, which are otherwise undesirable for memory application, boost the energy-efficiency of the GAN implementation on passive RRAM crossbar arrays without degrading its accuracy.
There has been immense development in the area of generative algorithms in recent years. Contrary to the discriminative models, which map high dimensional inputs to class labels, generative models are used in Variational autoencoders (VAEs) and Generative Adversarial Networks (GANs). Extensive studies have been done to improve unsupervised learning and GANs are one of the most successful algorithms to come up in the domain. With the benefits of providing greater accuracies, GANs have been expensive in terms of energy and speed, due to the vast number of Vector Matrix Multiplications computed on a large weight matrix. To overcome this hurdle, several works have been done on GPU and FPGA based accelerators. However, the Von Neumann bottleneck limits the accuracies and energy efficiency one can achieve and so Neuromorphic computing has been adopted greatly to exceed these limits. In this work, we have proposed an implementation of GANs on passive memristor crossbar arrays. We have performed a fixed amplitude training to update the weights with the crossbar as the backend. We also proposed to use a true random noise for the network. The simulation results show that our implementation has low energy consumption with comparable accuracies to the software counterpart.
Generative algorithms such as GANs are at the cusp of next revolution in the field of unsupervised learning and large-scale artificial data generation. However, the adversarial (competitive) co-training of the discriminative and generative networks in GAN makes them computationally intensive and hinders their deployment on the resource-constrained IoT edge devices. Moreover, the frequent data transfer between the discriminative and generative networks during training significantly degrades the efficacy of the von-Neumann GAN accelerators such as those based on GPU and FPGA. Therefore, there is an urgent need for development of ultra-compact and energy-efficient hardware accelerators for GANs. To this end, in this work, we propose to exploit the passive RRAM crossbar arrays for performing key operations of a fully-connected GAN: (a) true random noise generation for the generator network, (b) vector-by-matrix-multiplication with unprecedented energy-efficiency during the forward pass and backward propagation and (C) in-situ adversarial training using a hardware friendly Manhattan's rule. Our extensive analysis utilizing an experimentally calibrated phenomological model for passive RRAM crossbar array reveals an unforeseen trade-off between the accuracy and the energy dissipated while training the GAN network with different noise inputs to the generator. Furthermore, our results indicate that the spatial and temporal variations and true random noise, which are otherwise undesirable for memory application, boost the energy-efficiency of the GAN implementation on passive RRAM crossbar arrays without degrading its accuracy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.