Mirror Gradient Ascent Descent Algorithm

AbstractTask variance regularization, which can be used to improve the generalization of Multi-task Learning (MTL) models, remains unexplored in multi-task text classification. Accordingly, to fill this gap, this paper investigates …...