Towards Efficient Edge Learning with Limited Storage Resource: Bandit-Based Training Data Retrieval in AIoT

In an intelligent IoT environment, an edge server needs to retrieve data from end devices to train the deep neural network deployed at edge. In order to achieve higher training performance with limited resources, we propose a Bandit-based in-network training data retrieval scheme(Bandit-TDRetrieval). Specifically, we formulate data retrieval from end devices with a multi-armed bandit (MAB) model. A sequence of lever pulls of the arms, indicates the options to retrieve data from different end devices and follows a binomial distribution. To identify the relationship between this binomial distribution and the rewards through continuous data retrieval from the corresponding devices, Thompson sampling is used. In this way, an efficient data retrieval paradigm is proposed to maximize the rewards to retrieve meaningful training data for edge learning. Finally, the evaluations are performed on a NS-3-based simulation platform, which demonstrate the proposed paradigm effectively improve the training efficiency of the deep neural networks at edge.