[ASPLOS’25] Generalizing Reuse Patterns for Efficient DNN on Microcontrollers

Proposed Generalized Reuse, a framework that expands computation reuse strategies in neural networks, yielding 1.03-2.2× inference speedups or 1-8% accuracy improvements across diverse architectures.

Recommended citation: Jiesong Liu, Brian Park, Xipeng Shen. (2025). "Generalizing Reuse Patterns for Efficient DNN on Microcontrollers." ASPLOS 2025
Download Paper