Latent Domain Learning with Dynamic Residual Adapters
This addresses the challenge of domain-agnostic learning in practical scenarios lacking labeled domains, though it is incremental as it builds on existing domain adaptation and style transfer techniques.
The paper tackles the problem of learning from multi-domain data without domain annotations, where standard training overfits large domains and ignores smaller ones, and demonstrates that dynamic residual adapters significantly outperform larger off-the-shelf networks in robust performance across latent domains.
A practical shortcoming of deep neural networks is their specialization to a single task and domain. While recent techniques in domain adaptation and multi-domain learning enable the learning of more domain-agnostic features, their success relies on the presence of domain labels, typically requiring manual annotation and careful curation of datasets. Here we focus on a less explored, but more realistic case: learning from data from multiple domains, without access to domain annotations. In this scenario, standard model training leads to the overfitting of large domains, while disregarding smaller ones. We address this limitation via dynamic residual adapters, an adaptive gating mechanism that helps account for latent domains, coupled with an augmentation strategy inspired by recent style transfer techniques. Our proposed approach is examined on image classification tasks containing multiple latent domains, and we showcase its ability to obtain robust performance across these. Dynamic residual adapters significantly outperform off-the-shelf networks with much larger capacity, and can be incorporated seamlessly with existing architectures in an end-to-end manner.