MDLdroid:Mobile Deep Learning for Personal Mobile Sensing Applications

Personal mobile sensing is fast permeating our daily lives to enable activity monitoring, healthcare and rehabilitation. Towards pushing deep learning on devices, we present MDLdroid, our first version of a novel decentralized mobile deep learning framework to enable resource-aware on-device collaborative learning for personal mobile sensing applications. The framework leverages on multiple mobile devices connected via a mesh network to realize complete training and interference on devices. Our evaluations show that our model training on off-the-shelf mobile devices achieves 2x to 3.5x faster than single-device training, and 1.5x faster than the master-slave approach.

This work has been published in IPSN 2020 .