Training deep physical neural networks with local physical information bottleneck
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912894242783232 |
|---|---|
| author | Wang, Hao Wang, Ziao Liang, Xiangpeng Zhao, Han Hu, Jianqi Jiang, Junjie Fu, Xing Tang, Jianshi Wu, Huaqiang Gigan, Sylvain Liu, Qiang |
| author_facet | Wang, Hao Wang, Ziao Liang, Xiangpeng Zhao, Han Hu, Jianqi Jiang, Junjie Fu, Xing Tang, Jianshi Wu, Huaqiang Gigan, Sylvain Liu, Qiang |
| contents | Deep learning has revolutionized modern society but faces growing energy and latency constraints. Deep physical neural networks (PNNs) are interconnected computing systems that directly exploit analog dynamics for energy-efficient, ultrafast AI execution. Realizing this potential, however, requires universal training methods tailored to physical intricacies. Here, we present the Physical Information Bottleneck (PIB), a general and efficient framework that integrates information theory and local learning, enabling deep PNNs to learn under arbitrary physical dynamics. By allocating matrix-based information bottlenecks to each unit, we demonstrate supervised, unsupervised, and reinforcement learning across electronic memristive chips and optical computing platforms. PIB also adapts to severe hardware faults and allows for parallel training via geographically distributed resources. Bypassing auxiliary digital models and contrastive measurements, PIB recasts PNN training as an intrinsic, scalable information-theoretic process compatible with diverse physical substrates. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_09569 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Training deep physical neural networks with local physical information bottleneck Wang, Hao Wang, Ziao Liang, Xiangpeng Zhao, Han Hu, Jianqi Jiang, Junjie Fu, Xing Tang, Jianshi Wu, Huaqiang Gigan, Sylvain Liu, Qiang Machine Learning Applied Physics Deep learning has revolutionized modern society but faces growing energy and latency constraints. Deep physical neural networks (PNNs) are interconnected computing systems that directly exploit analog dynamics for energy-efficient, ultrafast AI execution. Realizing this potential, however, requires universal training methods tailored to physical intricacies. Here, we present the Physical Information Bottleneck (PIB), a general and efficient framework that integrates information theory and local learning, enabling deep PNNs to learn under arbitrary physical dynamics. By allocating matrix-based information bottlenecks to each unit, we demonstrate supervised, unsupervised, and reinforcement learning across electronic memristive chips and optical computing platforms. PIB also adapts to severe hardware faults and allows for parallel training via geographically distributed resources. Bypassing auxiliary digital models and contrastive measurements, PIB recasts PNN training as an intrinsic, scalable information-theoretic process compatible with diverse physical substrates. |
| title | Training deep physical neural networks with local physical information bottleneck |
| topic | Machine Learning Applied Physics |
| url | https://arxiv.org/abs/2602.09569 |