Cross-domain Few-shot Learning with Task-specific Adapters
This work addresses the problem of adapting models to new domains with limited data for researchers in few-shot learning, representing an incremental improvement over existing methods.
The paper tackles cross-domain few-shot classification by learning task-specific weights directly from a small support set using parametric adapters with residual connections, achieving significant performance improvements on the Meta-Dataset benchmark.
In this paper, we look at the problem of cross-domain few-shot classification that aims to learn a classifier from previously unseen classes and domains with few labeled samples. Recent approaches broadly solve this problem by parameterizing their few-shot classifiers with task-agnostic and task-specific weights where the former is typically learned on a large training set and the latter is dynamically predicted through an auxiliary network conditioned on a small support set. In this work, we focus on the estimation of the latter, and propose to learn task-specific weights from scratch directly on a small support set, in contrast to dynamically estimating them. In particular, through systematic analysis, we show that task-specific weights through parametric adapters in matrix form with residual connections to multiple intermediate layers of a backbone network significantly improves the performance of the state-of-the-art models in the Meta-Dataset benchmark with minor additional cost.