Joint Task and Resource Allocation for Mobile Edge Learning

The exploding increase in the number of connected devices and growing sizes of their generated data gave more opportunities for distributed learning to dominate fast data analytics in mobile edge environments. In this work, we aim to jointly optimize the allocation of learning tasks and wireless resources in such environments with the aim of maximizing the number of local training cycles each device executes within a given time constraint, which was shown to achieve a faster convergence to the desired learning accuracy. This joint problem is formulated as a non-linear constrained integer-linear problem, which is proven to be NP-hard. The problem is then simplified into a simpler form by deducing the optimal solution for some parameters. We then employ numerical solvers to efficiently solve this simplified problem. Simulation results show gains up to 166% and 250% compared to the task allocation only and the resource allocation only techniques, respectively.