Latent Domain Learning with Dynamic Residual Adapters
arXiv:2006.00996
Abstract
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.
References in corpus (13)
- mixup: Beyond Empirical Risk Minimization
- Neural Architecture Search with Reinforcement Learning
- Categorical Reparameterization with Gumbel-Softmax
- Efficient Neural Architecture Search via Parameter Sharing
- Learning multiple visual domains with residual adapters
- GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks
- Neural Architecture Search with Bayesian Optimisation and Optimal Transport
- AdaShare: Learning What To Share For Efficient Deep Multi-Task Learning
- Universal representations:The missing link between faces, text, planktons, and cat breeds
- Learning to Learn without Forgetting by Maximizing Transfer and Minimizing Interference
- Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift
- Episodic Training for Domain Generalization
- Towards Universal Object Detection by Domain Attention