Generalization Tower Network: A Novel Deep Neural Network Architecture for Multi-Task Learning
This addresses the challenge of learning multiple tasks efficiently in reinforcement learning, though it appears incremental as it builds on existing deep learning methods for RL.
The paper tackles the problem of multi-task reinforcement learning (MT-RL) by proposing a novel deep neural network architecture called generalization tower network (GTN), which uses horizontal and vertical streams to learn shared and hierarchical representations across tasks, achieving state-of-the-art results on 51 Atari games.
Deep learning (DL) advances state-of-the-art reinforcement learning (RL), by incorporating deep neural networks in learning representations from the input to RL. However, the conventional deep neural network architecture is limited in learning representations for multi-task RL (MT-RL), as multiple tasks can refer to different kinds of representations. In this paper, we thus propose a novel deep neural network architecture, namely generalization tower network (GTN), which can achieve MT-RL within a single learned model. Specifically, the architecture of GTN is composed of both horizontal and vertical streams. In our GTN architecture, horizontal streams are used to learn representation shared in similar tasks. In contrast, the vertical streams are introduced to be more suitable for handling diverse tasks, which encodes hierarchical shared knowledge of these tasks. The effectiveness of the introduced vertical stream is validated by experimental results. Experimental results further verify that our GTN architecture is able to advance the state-of-the-art MT-RL, via being tested on 51 Atari games.