Neuromorphic Photonic Computing with an Electro-Optic Analog Memory

2024-01-29

AI Summary

Conventional digital computing faces fundamental energy-efficiency limitations, and AI workloads are increasingly constrained by computational power and energy bottlenecks. Method: This work proposes a monolithically integrated neuromorphic photonic computing architecture, co-integrating capacitive electro-optic analog memory with photonic computing units on a single chip—enabling in-situ analog-domain computation and eliminating repeated digital-to-analog (DAC) and analog-to-digital (ADC) conversions as well as costly data movement. Contribution/Results: It reports the first monolithic co-integration of capacitive electro-optic analog memory with photonic devices and establishes a comprehensive performance metric framework for analog memory tailored to neuromorphic photonic computing. Evaluated on MNIST, the architecture reduces DAC dependency and memory access energy significantly; its computational energy efficiency improves by over one order of magnitude compared to conventional digital approaches, offering a scalable hardware paradigm for energy-efficient neuromorphic photonic computing.

Technology Category

• Machine Learning: Hardware-aware ML
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Learning & Optimization for CV

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Artificial intelligence (AI) has seen remarkable advancements across various domains, including natural language processing, computer vision, autonomous vehicles, and biology. However, the rapid expansion of AI technologies has escalated the demand for more powerful computing resources. As digital computing approaches fundamental limits, neuromorphic photonics emerges as a promising platform to complement existing digital systems. In neuromorphic photonic computing, photonic devices are controlled using analog signals. This necessitates the use of digital-to-analog converters (DAC) and analog-to-digital converters (ADC) for interfacing with these devices during inference and training. However, data movement between memory and these converters in conventional von Neumann computing architectures consumes energy. To address this, analog memory co-located with photonic computing devices is proposed. This approach aims to reduce the reliance on DACs and minimize data movement to enhance compute efficiency. This paper demonstrates a monolithically integrated neuromorphic photonic circuit with co-located capacitive analog memory and analyzes analog memory specifications for neuromorphic photonic computing using the MNIST dataset as a benchmark.

Problem

• Addresses energy inefficiency in neuromorphic photonic computing systems.
• Proposes analog memory co-location to reduce DAC and ADC reliance.
• Demonstrates integrated photonic circuit with analog memory using MNIST.

Innovation

• Monolithic integration of photonic circuits
• Co-located capacitive analog memory
• Reduced reliance on DACs and ADCs