Clustered Task-Aware Meta-Learning by Learning from Learning Paths

To enable effective learning of new tasks with only few samples, meta-learning acquires common knowledge from the existing tasks with a globally shared meta-learner. To further address the problem of task heterogeneity, recent developments balance between customization and generalization by incorporating task clustering to generate the task-aware modulation to be applied on the global meta-learner. However, these methods learn task representation mostly from the features of input data, while the task-specific optimization process with respect to the base-learner model is often neglected. In this work, we propose a Clustered Task-Aware Meta-Learning (CTML) framework with task representation learned from its own learning path. We first conduct a rehearsed task learning from the common initialization, and collect a set of geometric quantities that adequately describes this learning path. By inputting this set of values into a meta path learner, we automatically abstract path representation optimized for the downstream clustering and modulation. To further save the computational cost incurred by the additional rehearsed learning, we devise a shortcut tunnel to directly map between the path and feature cluster assignments. Extensive experiments on two real-world application domains: few-shot image classification and cold-start recommendation demonstrate the superiority of CTML compared to state-of-the-art baselines.