Auction-based Deep Learning Computation Offloading for Truthful Edge Computing: A Myerson Auction Approach

Abstract

This paper proposes an edge computing model that processes machine learning code offloading based on auction mechanisms. The code offloading is required to carry out training that is processed difficult for mobile devices with limited computing resources. In this system, mobile devices compete for using computing resource units by submitting their bids based on code complexity and data size. We use the Myerson auction model which uses the truthful second-price auction as a baseline, to maximize the seller's revenue while meeting several desirable properties, i.e., individual rationality and incentive compatibility. The simulation results showed that our Myerson auction method overall improved the seller's revenue while satisfying the above characteristics.