Hierarchical Meta Learning
This addresses the inefficiency of meta learning for heterogeneous few-shot learning tasks, making it more applicable in real-world scenarios where task structures vary.
The paper tackles the problem of meta learning being limited to tasks with the same output structure, which requires retraining for new structures, by developing a Hierarchical Meta Learning (HML) method that optimizes both adaptability and generalizability across heterogeneous tasks, resulting in superior performance in few-shot classification and regression compared to existing approaches.
Meta learning is a promising solution to few-shot learning problems. However, existing meta learning methods are restricted to the scenarios where training and application tasks share the same out-put structure. To obtain a meta model applicable to the tasks with new structures, it is required to collect new training data and repeat the time-consuming meta training procedure. This makes them inefficient or even inapplicable in learning to solve heterogeneous few-shot learning tasks. We thus develop a novel and principled HierarchicalMeta Learning (HML) method. Different from existing methods that only focus on optimizing the adaptability of a meta model to similar tasks, HML also explicitly optimizes its generalizability across heterogeneous tasks. To this end, HML first factorizes a set of similar training tasks into heterogeneous ones and trains the meta model over them at two levels to maximize adaptation and generalization performance respectively. The resultant model can then directly generalize to new tasks. Extensive experiments on few-shot classification and regression problems clearly demonstrate the superiority of HML over fine-tuning and state-of-the-art meta learning approaches in terms of generalization across heterogeneous tasks.