Paper Info
Title
ADEPT: automatic differentiable DEsign of photonic tensor cores
Abstract
Photonic tensor cores (PTCs) are essential building blocks for optical artificial intelligence (AI) accelerators based on programmable photonic integrated circuits. PTCs can achieve ultra-fast and efficient tensor operations for neural network (NN) acceleration. Current PTC designs are either manually constructed or based on matrix decomposition theory, which lacks the adaptability to meet various hardware constraints and device specifications. To our best knowledge, automatic PTC design methodology is still unexplored. It will be promising to move beyond the manual design paradigm and "nurture" photonic neurocomputing with AI and design automation. Therefore, in this work, for the first time, we propose a fully differentiable framework, dubbed ADEPT, that can efficiently search PTC designs adaptive to various circuit footprint constraints and foundry PDKs. Extensive experiments show superior flexibility and effectiveness of the proposed ADEPT framework to explore a large PTC design space. On various NN models and benchmarks, our searched PTC topology outperforms prior manually-designed structures with competitive matrix representability, 2x-30x higher footprint compactness, and better noise robustness, demonstrating a new paradigm in photonic neural chip design.
Year
DOI
Venue
2022
10.1145/3489517.3530562
Design Automation Conference (DAC)
DocType
Citations 
PageRank 
Conference
0
0.34
References 
Authors
0
8
Name
Order
Citations
PageRank
Jiaqi Gu1186.97
Hanqing Zhu201.35
Chenghao Feng353.96
Zixuan Jiang472.49
Mingjie Liu517819.18
Shuhan Zhang6106.28
T. Chen74910.96
David Z. Pan82653237.64