ELight: Enabling Efficient Photonic In-Memory Neurocomputing with Life Enhancement
This work addresses efficiency and reliability issues for photonic in-memory neurocomputing systems, enabling longer lifetime and lower energy in machine learning applications, though it is incremental as it optimizes an existing approach.
The paper tackles the problem of high dynamic power and limited write endurance in photonic in-memory neurocomputing due to massive hardware reuse, proposing ELight, a synergistic optimization framework that reduces the total number of writes and dynamic power by over 20X while maintaining comparable accuracy.
With the recent advances in optical phase change material (PCM), photonic in-memory neurocomputing has demonstrated its superiority in optical neural network (ONN) designs with near-zero static power consumption, time-of-light latency, and compact footprint. However, photonic tensor cores require massive hardware reuse to implement large matrix multiplication due to the limited single-core scale. The resultant large number of PCM writes leads to serious dynamic power and overwhelms the fragile PCM with limited write endurance. In this work, we propose a synergistic optimization framework, ELight, to minimize the overall write efforts for efficient and reliable optical in-memory neurocomputing. We first propose write-aware training to encourage the similarity among weight blocks, and combine it with a post-training optimization method to reduce programming efforts by eliminating redundant writes. Experiments show that ELight can achieve over 20X reduction in the total number of writes and dynamic power with comparable accuracy. With our ELight, photonic in-memory neurocomputing will step forward towards viable applications in machine learning with preserved accuracy, order-of-magnitude longer lifetime, and lower programming energy.