Training deep physical neural networks with local physical information bottleneck
This addresses the energy and latency constraints in deep learning for AI applications, though it is incremental as it builds on existing PNN concepts with a novel training method.
The paper tackles the challenge of training deep physical neural networks (PNNs) for energy-efficient AI by introducing the Physical Information Bottleneck (PIB) framework, which enables supervised, unsupervised, and reinforcement learning across electronic and optical platforms without auxiliary digital models.
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.