Learning Centric Wireless Resource Allocation for Edge Computing: Algorithm and Experiment
This addresses a fundamental communication challenge in edge intelligence for machine learning applications, though it appears incremental by focusing on specific overlooked factors in resource allocation.
The paper tackles the problem of allocating limited wireless resources for simultaneous model training of heterogeneous learning tasks in edge computing, proposing a learning-centric scheme that maximizes worst-case learning performance and showing an inverse power relationship between transmission time and generalization error, with simulation and experimental results verifying its performance and robustness.
Edge intelligence is an emerging network architecture that integrates sensing, communication, computing components, and supports various machine learning applications, where a fundamental communication question is: how to allocate the limited wireless resources (such as time, energy) to the simultaneous model training of heterogeneous learning tasks? Existing methods ignore two important facts: 1) different models have heterogeneous demands on training data; 2) there is a mismatch between the simulated environment and the real-world environment. As a result, they could lead to low learning performance in practice. This paper proposes the learning centric wireless resource allocation (LCWRA) scheme that maximizes the worst learning performance of multiple tasks. Analysis shows that the optimal transmission time has an inverse power relationship with respect to the generalization error. Finally, both simulation and experimental results are provided to verify the performance of the proposed LCWRA scheme and its robustness in real implementation.