Adaptation-Agnostic Meta-Training

TL;DR

This paper introduces an adaptation-agnostic meta-training approach that allows the use of complex, non-analytical inner-task algorithms, enhancing flexibility and performance in meta-learning.

Contribution

It proposes a novel meta-training strategy that removes the need for analytical solutions in inner-task algorithms, enabling the use of more powerful and diverse methods.

Findings

Enables application of complex inner-task algorithms

Achieves superior performance over traditional meta-learning methods

Supports ensemble of different algorithm types

Abstract

Many meta-learning algorithms can be formulated into an interleaved process, in the sense that task-specific predictors are learned during inner-task adaptation and meta-parameters are updated during meta-update. The normal meta-training strategy needs to differentiate through the inner-task adaptation procedure to optimize the meta-parameters. This leads to a constraint that the inner-task algorithms should be solved analytically. Under this constraint, only simple algorithms with analytical solutions can be applied as the inner-task algorithms, limiting the model expressiveness. To lift the limitation, we propose an adaptation-agnostic meta-training strategy. Following our proposed strategy, we can apply stronger algorithms (e.g., an ensemble of different types of algorithms) as the inner-task algorithm to achieve superior performance comparing with popular baselines. The source code is available at https://github.com/jiaxinchen666/AdaptationAgnosticMetaLearning.