Optimization Framework for Splitting DNN Inference Jobs over Computing Networks

Abstract

Ubiquitous artificial intelligence (AI) is considered one of the key services in 6G systems. AI services typically rely on deep neural network (DNN) requiring heavy computation. Hence, in order to support ubiquitous AI, it is crucial to provide a solution for offloading or distributing computational burden due to DNN, especially at end devices with limited resources. We develop a framework for assigning the computation tasks of DNN inference jobs to the nodes with computing resources in the network, so as to reduce the inference latency in the presence of limited computing power at end devices. To this end, we propose a layered graph model that enables to solve the problem of assigning computation tasks of a single DNN inference job via simple conventional routing. Using this model, we develop algorithms for routing DNN inference jobs over the distributed computing network. We show through numerical evaluations that our algorithms can select nodes and paths adaptively to the computational attributes of given DNN inference jobs in order to reduce the end-to-end latency.