AdaBridge: Dynamic Data and Computation Reuse for Efficient Multi-task DNN Co-evolution in Edge Systems

TL;DR

AdaBridge enhances multi-task DNN co-evolution on edge devices by exploiting redundancies for dynamic data and computation reuse, leading to improved accuracy and resource efficiency.

Contribution

It introduces AdaBridge, a novel method for dynamic data and computation reuse in multi-task DNN evolution on edge systems, addressing a key open challenge.

Findings

Achieves 11% average accuracy gain over baseline methods.

Improves training efficiency and resource utilization.

Effectively handles asynchronous multi-task co-evolution.

Abstract

Running multi-task DNNs on mobiles is an emerging trend for various applications like autonomous driving and mobile NLP. Mobile DNNs are often compressed to fit the limited resources and thus suffer from degraded accuracy and generalizability due to data drift. DNN evolution, e.g., continuous learning and domain adaptation, has been demonstrated effective in overcoming these issues, mostly for single-task DNN, leaving multi-task DNN evolution an important yet open challenge. To fill up this gap, we propose AdaBridge, which exploits computational redundancies in multi-task DNNs as a unique opportunity for dynamic data and computation reuse, thereby improving training efficacy and resource efficiency among asynchronous multi-task co-evolution in edge systems. Experimental evaluation shows that AdaBridge achieves 11% average accuracy gain upon individual evolution baselines.