PhotoGAN: Generative Adversarial Neural Network Acceleration with Silicon Photonics

2025-01-23

AI Summary

To address the low energy efficiency and high computational overhead of Generative Adversarial Networks (GANs) on conventional electronic accelerators, this work proposes the first silicon-photonics-based reconfigurable accelerator tailored for GANs. Focusing on key inefficient operators—particularly transposed convolution and instance normalization—we pioneer the integration of silicon photonics into GAN-specific acceleration, designing a reconfigurable optical-domain architecture. Our approach jointly incorporates sparse tensor computation optimization and a novel optical-domain mapping strategy for transposed convolution. Implemented using a silicon photonic integrated circuit and a programmable optical interferometric array, the prototype maintains full model accuracy while achieving a 4.4× improvement in GOPPS and a 2.18× reduction in energy per byte (EPB) over state-of-the-art GPU/TPU platforms. This work breaks the energy-efficiency bottleneck inherent in electronic hardware and establishes a scalable, reconfigurable paradigm for photonic acceleration of AI models.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Hardware-aware ML
• Search and Optimization: Non-convex Optimization

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Generative Adversarial Networks (GANs) are at the forefront of AI innovation, driving advancements in areas such as image synthesis, medical imaging, and data augmentation. However, the unique computational operations within GANs, such as transposed convolutions and instance normalization, introduce significant inefficiencies when executed on traditional electronic accelerators, resulting in high energy consumption and suboptimal performance. To address these challenges, we introduce PhotoGAN, the first silicon-photonic accelerator designed to handle the specialized operations of GAN models. By leveraging the inherent high throughput and energy efficiency of silicon photonics, PhotoGAN offers an innovative, reconfigurable architecture capable of accelerating transposed convolutions and other GAN-specific layers. The accelerator also incorporates a sparse computation optimization technique to reduce redundant operations, improving computational efficiency. Our experimental results demonstrate that PhotoGAN achieves at least 4.4x higher GOPS and 2.18x lower energy-per-bit (EPB) compared to state-of-the-art accelerators, including GPUs and TPUs. These findings showcase PhotoGAN as a promising solution for the next generation of GAN acceleration, providing substantial gains in both performance and energy efficiency.

Problem

• Energy Efficiency
• Generative Adversarial Networks (GANs)
• Image Generation

Innovation

• PhotoGAN
• Light-based Acceleration
• Energy-efficient GANs