Few-Shot Adversarial Domain Adaptation

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.

Accepted to NIPS 2017. arXiv admin note: text overlap with arXiv:1709.10190

References in corpus (10)

• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Deep Domain Confusion: Maximizing for Domain Invariance
• Improved Techniques for Training GANs
• NIPS 2016 Tutorial: Generative Adversarial Networks
• Beyond Sharing Weights for Deep Domain Adaptation
• Adversarial Discriminative Domain Adaptation
• Deep CORAL: Correlation Alignment for Deep Domain Adaptation
• Learning Using Privileged Information: SVM+ and Weighted SVM
• Domain Adaptation by Mixture of Alignments of Second- or Higher-Order Scatter Tensors

Cited by in corpus (25)

• Deep Visual Domain Adaptation: A Survey
• COVID-19 Image Data Collection: Prospective Predictions Are the Future
• Multi-Site Infant Brain Segmentation Algorithms: The iSeg-2019 Challenge
• Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift
• Supervised Domain Adaptation: A Graph Embedding Perspective and a Rectified Experimental Protocol
• Partial Adversarial Domain Adaptation
• Adaptive Betweenness Clustering for Semi-Supervised Domain Adaptation
• PhyAug: Physics-Directed Data Augmentation for Deep Sensing Model Transfer in Cyber-Physical Systems
• Scaling up sign spotting through sign language dictionaries
• Overview of LifeCLEF Plant Identification task 2020
• Few shot domain adaptation for in situ macromolecule structural classification in cryo-electron tomograms
• Cryo-shift: Reducing domain shift in cryo-electron subtomograms with unsupervised domain adaptation and randomization
• Few-Shot Learning as Domain Adaptation: Algorithm and Analysis
• Few-shot Unsupervised Domain Adaptation with Image-to-class Sparse Similarity Encoding
• Exploiting Local Feature Patterns for Unsupervised Domain Adaptation
• Unsupervised Domain Adaptation across FMCW Radar Configurations Using Margin Disparity Discrepancy
• Annotation Cost Reduction of Stream-based Active Learning by Automated Weak Labeling using a Robot Arm
• Sequential Sampling for Optimal Bayesian Classification of Sequencing Count Data
• One-shot Unsupervised Domain Adaptation with Personalized Diffusion Models
• Towards Robust Pattern Recognition: A Review
• MetaXT: Meta Cross-Task Transfer between Disparate Label Spaces
• Deep Cross-modality Adaptation via Semantics Preserving Adversarial Learning for Sketch-based 3D Shape Retrieval
• One-shot domain adaptation for semantic face editing of real world images using StyleALAE
• Task Guided Compositional Representation Learning for ZDA
• Collaborating Foundation Models for Domain Generalized Semantic Segmentation