Attentive Task Interaction Network for Multi-Task Learning
This work addresses feature sharing for multitask learning practitioners, offering an incremental improvement over existing methods.
The paper tackles the challenge of feature sharing in multitask learning by proposing the Attentive Task Interaction Network (ATI-Net), which uses knowledge distillation and attention to combine features, resulting in improved performance over state-of-the-art baselines like MTAN and PAD-Net with similar model parameters.
Multitask learning (MTL) has recently gained a lot of popularity as a learning paradigm that can lead to improved per-task performance while also using fewer per-task model parameters compared to single task learning. One of the biggest challenges regarding MTL networks involves how to share features across tasks. To address this challenge, we propose the Attentive Task Interaction Network (ATI-Net). ATI-Net employs knowledge distillation of the latent features for each task, then combines the feature maps to provide improved contextualized information to the decoder. This novel approach to introducing knowledge distillation into an attention based multitask network outperforms state of the art MTL baselines such as the standalone MTAN and PAD-Net, with roughly the same number of model parameters.