Photonic neuromorphic architecture for tens-of-task lifelong learning
Abstract Scalable, high-capacity, and low-power computing architecture is the primary assurance for increasingly manifold and large-scale machine learning tasks. Traditional electronic artificial agents by conventional power-hungry processors have faced the issues of energy and scaling walls, hinder...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Publishing Group
2024-02-01
|
| Series: | Light: Science & Applications |
| Online Access: | https://doi.org/10.1038/s41377-024-01395-4 |
| _version_ | 1797273168250929152 |
|---|---|
| author | Yuan Cheng Jianing Zhang Tiankuang Zhou Yuyan Wang Zhihao Xu Xiaoyun Yuan Lu Fang |
| author_facet | Yuan Cheng Jianing Zhang Tiankuang Zhou Yuyan Wang Zhihao Xu Xiaoyun Yuan Lu Fang |
| author_sort | Yuan Cheng |
| collection | DOAJ |
| description | Abstract Scalable, high-capacity, and low-power computing architecture is the primary assurance for increasingly manifold and large-scale machine learning tasks. Traditional electronic artificial agents by conventional power-hungry processors have faced the issues of energy and scaling walls, hindering them from the sustainable performance improvement and iterative multi-task learning. Referring to another modality of light, photonic computing has been progressively applied in high-efficient neuromorphic systems. Here, we innovate a reconfigurable lifelong-learning optical neural network (L2ONN), for highly-integrated tens-of-task machine intelligence with elaborated algorithm-hardware co-design. Benefiting from the inherent sparsity and parallelism in massive photonic connections, L2ONN learns each single task by adaptively activating sparse photonic neuron connections in the coherent light field, while incrementally acquiring expertise on various tasks by gradually enlarging the activation. The multi-task optical features are parallelly processed by multi-spectrum representations allocated with different wavelengths. Extensive evaluations on free-space and on-chip architectures confirm that for the first time, L2ONN avoided the catastrophic forgetting issue of photonic computing, owning versatile skills on challenging tens-of-tasks (vision classification, voice recognition, medical diagnosis, etc.) with a single model. Particularly, L2ONN achieves more than an order of magnitude higher efficiency than the representative electronic artificial neural networks, and 14× larger capacity than existing optical neural networks while maintaining competitive performance on each individual task. The proposed photonic neuromorphic architecture points out a new form of lifelong learning scheme, permitting terminal/edge AI systems with light-speed efficiency and unprecedented scalability. |
| first_indexed | 2024-03-07T14:39:43Z |
| format | Article |
| id | doaj.art-64955bb6c13744b2a6c908bb954514ea |
| institution | Directory Open Access Journal |
| issn | 2047-7538 |
| language | English |
| last_indexed | 2024-03-07T14:39:43Z |
| publishDate | 2024-02-01 |
| publisher | Nature Publishing Group |
| record_format | Article |
| series | Light: Science & Applications |
| spelling | doaj.art-64955bb6c13744b2a6c908bb954514ea2024-03-05T20:24:35ZengNature Publishing GroupLight: Science & Applications2047-75382024-02-0113111210.1038/s41377-024-01395-4Photonic neuromorphic architecture for tens-of-task lifelong learningYuan Cheng0Jianing Zhang1Tiankuang Zhou2Yuyan Wang3Zhihao Xu4Xiaoyun Yuan5Lu Fang6Sigma Laboratory, Department of Electronic Engineering, Tsinghua UniversitySigma Laboratory, Department of Electronic Engineering, Tsinghua UniversitySigma Laboratory, Department of Electronic Engineering, Tsinghua UniversityBeijing National Research Center for Information Science and Technology (BNRist)Sigma Laboratory, Department of Electronic Engineering, Tsinghua UniversitySigma Laboratory, Department of Electronic Engineering, Tsinghua UniversitySigma Laboratory, Department of Electronic Engineering, Tsinghua UniversityAbstract Scalable, high-capacity, and low-power computing architecture is the primary assurance for increasingly manifold and large-scale machine learning tasks. Traditional electronic artificial agents by conventional power-hungry processors have faced the issues of energy and scaling walls, hindering them from the sustainable performance improvement and iterative multi-task learning. Referring to another modality of light, photonic computing has been progressively applied in high-efficient neuromorphic systems. Here, we innovate a reconfigurable lifelong-learning optical neural network (L2ONN), for highly-integrated tens-of-task machine intelligence with elaborated algorithm-hardware co-design. Benefiting from the inherent sparsity and parallelism in massive photonic connections, L2ONN learns each single task by adaptively activating sparse photonic neuron connections in the coherent light field, while incrementally acquiring expertise on various tasks by gradually enlarging the activation. The multi-task optical features are parallelly processed by multi-spectrum representations allocated with different wavelengths. Extensive evaluations on free-space and on-chip architectures confirm that for the first time, L2ONN avoided the catastrophic forgetting issue of photonic computing, owning versatile skills on challenging tens-of-tasks (vision classification, voice recognition, medical diagnosis, etc.) with a single model. Particularly, L2ONN achieves more than an order of magnitude higher efficiency than the representative electronic artificial neural networks, and 14× larger capacity than existing optical neural networks while maintaining competitive performance on each individual task. The proposed photonic neuromorphic architecture points out a new form of lifelong learning scheme, permitting terminal/edge AI systems with light-speed efficiency and unprecedented scalability.https://doi.org/10.1038/s41377-024-01395-4 |
| spellingShingle | Yuan Cheng Jianing Zhang Tiankuang Zhou Yuyan Wang Zhihao Xu Xiaoyun Yuan Lu Fang Photonic neuromorphic architecture for tens-of-task lifelong learning Light: Science & Applications |
| title | Photonic neuromorphic architecture for tens-of-task lifelong learning |
| title_full | Photonic neuromorphic architecture for tens-of-task lifelong learning |
| title_fullStr | Photonic neuromorphic architecture for tens-of-task lifelong learning |
| title_full_unstemmed | Photonic neuromorphic architecture for tens-of-task lifelong learning |
| title_short | Photonic neuromorphic architecture for tens-of-task lifelong learning |
| title_sort | photonic neuromorphic architecture for tens of task lifelong learning |
| url | https://doi.org/10.1038/s41377-024-01395-4 |
| work_keys_str_mv | AT yuancheng photonicneuromorphicarchitecturefortensoftasklifelonglearning AT jianingzhang photonicneuromorphicarchitecturefortensoftasklifelonglearning AT tiankuangzhou photonicneuromorphicarchitecturefortensoftasklifelonglearning AT yuyanwang photonicneuromorphicarchitecturefortensoftasklifelonglearning AT zhihaoxu photonicneuromorphicarchitecturefortensoftasklifelonglearning AT xiaoyunyuan photonicneuromorphicarchitecturefortensoftasklifelonglearning AT lufang photonicneuromorphicarchitecturefortensoftasklifelonglearning |