Mirage: An RNS-Based Photonic Accelerator for DNN Training

Abstract

Photonic computing is a compelling avenue for performing highly efficient matrix multiplication, a crucial operation in Deep Neural Networks (DNNs). While this method has shown great success in DNN inference, meeting the high precision demands of DNN training proves challenging due to the precision limitations imposed by costly data converters and the analog noise inherent in photonic hardware. This paper proposes Mirage, a photonic DNN training accelerator that overcomes the precision challenges in photonic hardware using the Residue Number System (RNS). RNS is a numeral system based on modular arithmetic, allowing us to perform high-precision operations via multiple low-precision modular operations. In this work, we present a novel micro-architecture and dataflow for an RNS-based photonic tensor core performing modular arithmetic in the analog domain. By combining RNS and photonics, Mirage provides high energy efficiency without compromising precision and can successfully train state-of-the-art DNNs achieving accuracy comparable to FP32 training. Our study shows that on average across several DNNs when compared to systolic arrays, Mirage achieves more than $23.8\times$ faster training and $32.1\times$ lower EDP in an iso-energy scenario and consumes $42.8\times$ lower power with comparable or better EDP in an iso-area scenario.

References in corpus (16)

• Deep Learning with Coherent Nanophotonic Circuits
• DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients
• 11 TeraFLOPs per second photonic convolutional accelerator for deep learning optical neural networks
• Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations
• Training of photonic neural networks through in situ backpropagation
• Experimentally realized in situ backpropagation for deep learning in nanophotonic neural networks
• Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation
• Scalable Methods for 8-bit Training of Neural Networks
• Hardware error correction for programmable photonics
• Silicon Photonic Architecture for Training Deep Neural Networks with Direct Feedback Alignment
• PCNNA: A Photonic Convolutional Neural Network Accelerator
• An Electro-Photonic System for Accelerating Deep Neural Networks
• A Modulo-Based Architecture for Analog-to-Digital Conversion
• Stability of Self-Configuring Large Multiport Interferometers
• A Blueprint for Precise and Fault-Tolerant Analog Neural Networks
• Adaptive Block Floating-Point for Analog Deep Learning Hardware