Title: A Brief Review of Deep Multi-task Learning and Auxiliary Task Learning

Abstract: Multi-task learning (MTL) optimizes several learning tasks simultaneously and leverages their shared information to improve generalization and the prediction of the model for each task. Auxiliary tasks can be added to the main task to ultimately boost the performance. In this paper, we provide a brief review on the recent deep multi-task learning (dMTL) approaches followed by methods on selecting useful auxiliary tasks that can be used in dMTL to improve the performance of the model for the main task.