ISCA: Intelligent Sense-Compute Adaptive Co-Optimization of Multimodal Machine Learning Kernels for Resilient mHealth Services on Wearables
Hamidreza Alikhani, Anil Kanduri, Emad Kasaeyan Naeini, Sina Shahhosseini, Pasi Liljeberg, Amir Masoud Rahmani, Nikil D. Dutt · IEEE Design and Test · 2024
Editor’s notes: This article proposes a sensing and computing adaptive optimization of multimodal machine learning kernels. The framework employs reinforcement learning to jointly determine sensing and compute configuration settings, which minimizes energy consumption while providing accuracy guarantees. The solution achieves up to 23% improvement in prediction accuracy compared to prior methods. —Ganapati Bhat, Washington State University, USA —Pietro Mercati, Intel Labs, USA