Exploiting Domain-Specific Features to Enhance Domain Generalization
Dec 6, 2021
Speakers
Manh-Ha Bui
Speaker · 0 followers
Dinh Phung
Speaker · 0 followers
About
Domain generalization (DG) aims to train a model, from multiple observed source domains, in order to perform well on unseen target domains. To obtain the generalization ability, prior DG approaches have focused on extracting Domain Invariant (DI) information across sources to generalize on target domains, while useful Domain Specific (DS) information which strongly correlates with labels in individual domains and the generalization to target domains is usually ignored. To this end, we propose in this paper meta-DomainSpecific-Domain Invariant (mDSDI) - a novel theoretically sound framework that extends beyond invariance view to further capture the usefulness of DS information. Our key insight is to disentangle features in the latent space while jointly learning both DI and DS features in a unified framework. The DS representation is optimized though the meta-learning framework to adapt from source domains, targeting a robust generalization on unseen domains. We empirically show that mDSDI provides competitive results with state-of-the-art techniques in DG. A further ablation study with our generated dataset, Colored-MNIST, confirms the hypothesis that DS is essential, leading to better results when compared with only using DI.Domain generalization (DG) aims to train a model, from multiple observed source domains, in order to perform well on unseen target domains. To obtain the generalization ability, prior DG approaches have focused on extracting Domain Invariant (DI) information across sources to generalize on target domains, while useful Domain Specific (DS) information which strongly correlates with labels in individual domains and the generalization to target domains is usually ignored. To this end, we propose in…
Organizer
NeurIPS 2021
Account · 1.9k followers
About NeurIPS 2021
Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference that includes invited talks, demonstrations, symposia and oral and poster presentations of refereed papers. Following the conference, there are workshops which provide a less formal setting.
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Aboriginal Turing Test
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Bridging the gap between AI and the life sciences: towards a standardized multi-omics data type
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Learning Adaptive Control Flow in Transformers for Improved Systematic Generalization
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Opening Remarks & Introduction
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Dataset Distillation with Infinitely Wide Convolutional Networks
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Learning Generalized Gumbel-max Causal Mechanisms
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%