Two Teachers Better Than One: Hardware-Physics Co-Guided Distributed Scientific Machine Learning

This paper presents EPIC, a hardware- and physics-co-guided distributed scientific machine learning framework for full-waveform inversion (FWI). The framework introduces edge-side encoding and physics-aware decoding to reduce communication latency and energy consumption while preserving physical fidelity in distributed scientific computing environments.

Recommended citation: Yuchen Yuan, Junhuan Yang, Hao Wan, Yipei Liu, Hanhan Wu, Youzuo Lin and Lei Yang. (2026). "Two Teachers Better Than One: Hardware-Physics Co-Guided Distributed Scientific Machine Learning." Proceedings of the 63rd Design Automation Conference (DAC).