Few-Shot Adversarial Domain Adaptation

TLDR

Supervised domain adaptation is attractive when only a few target samples are labeled, but few‑shot settings make aligning and separating semantic distributions difficult due to data scarcity. This work proposes a framework for supervised domain adaptation using deep models. The method employs adversarial learning to create an embedded subspace that maximizes domain confusion while aligning embeddings, and is evaluated against state‑of‑the‑art on handwritten digit and visual object datasets. Augmenting the adversarial discriminator to distinguish four classes enables effective supervised adaptation and achieves rapid adaptation with as few as one labeled sample per category.

Abstract

This work provides a framework for addressing the problem of supervised domain adaptation with deep models. The main idea is to exploit adversarial learning to learn an embedded subspace that simultaneously maximizes the confusion between two domains while semantically aligning their embedding. The supervised setting becomes attractive especially when there are only a few target data samples that need to be labeled. In this few-shot learning scenario, alignment and separation of semantic probability distributions is difficult because of the lack of data. We found that by carefully designing a training scheme whereby the typical binary adversarial discriminator is augmented to distinguish between four different classes, it is possible to effectively address the supervised adaptation problem. In addition, the approach has a high “speed” of adaptation, i.e. it requires an extremely low number of labeled target training samples, even one per category can be effective. We then extensively compare this approach to the state of the art in domain adaptation in two experiments: one using datasets for handwritten digit recognition, and one using datasets for visual object recognition.