Learning to Propagate Labels: Transductive Propagation Network for Few-shot Learning

Abstract

The goal of few-shot learning is to learn a classifier that generalizes well even when trained with a limited number of training instances per class. The recently introduced meta-learning approaches tackle this problem by learning a generic classifier across a large number of multiclass classification tasks and generalizing the model to a new task. Yet, even with such meta-learning, the low-data problem in the novel classification task still remains. In this paper, we propose Transductive Propagation Network (TPN), a novel meta-learning framework for transductive inference that classifies the entire test set at once to alleviate the low-data problem. Specifically, we propose to learn to propagate labels from labeled instances to unlabeled test instances, by learning a graph construction module that exploits the manifold structure in the data. TPN jointly learns both the parameters of feature embedding and the graph construction in an end-to-end manner. We validate TPN on multiple benchmark datasets, on which it largely outperforms existing few-shot learning approaches and achieves the state-of-the-art results.

arXiv commentAccepted in ICLR 2019; code available at https://github.com/csyanbin/TPN

Cited by in corpus (111)

• Generalizing from a Few Examples: A Survey on Few-Shot Learning
• Meta-Learning with Latent Embedding Optimization
• Low-Rank Pairwise Alignment Bilinear Network For Few-Shot Fine-Grained Image Classification
• A Two-Stage Approach to Few-Shot Learning for Image Recognition
• An Ensemble of Epoch-wise Empirical Bayes for Few-shot Learning
• PANet: Few-Shot Image Semantic Segmentation with Prototype Alignment
• Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification
• Unifying Graph Convolutional Neural Networks and Label Propagation
• TOAN: Target-Oriented Alignment Network for Fine-Grained Image Categorization with Few Labeled Samples
• Self-Regularized Prototypical Network for Few-Shot Semantic Segmentation
• Few-Shot Learning via Embedding Adaptation with Set-to-Set Functions
• PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences
• Learning to detect an animal sound from five examples
• Multi-level Second-order Few-shot Learning
• Transductive Information Maximization For Few-Shot Learning
• DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning
• A Comprehensive Overview and Survey of Recent Advances in Meta-Learning
• Transductive Episodic-Wise Adaptive Metric for Few-Shot Learning
• MHFC: Multi-Head Feature Collaboration for Few-Shot Learning
• Information Symmetry Matters: A Modal-Alternating Propagation Network for Few-Shot Learning
• Learning to Extrapolate Knowledge: Transductive Few-shot Out-of-Graph Link Prediction
• Leveraging the Feature Distribution in Transfer-based Few-Shot Learning
• TaskNorm: Rethinking Batch Normalization for Meta-Learning
• Meta-Learned Confidence for Few-shot Learning
• GCT: Graph Co-Training for Semi-Supervised Few-Shot Learning
• Part-aware Prototype Network for Few-shot Semantic Segmentation
• DPGN: Distribution Propagation Graph Network for Few-shot Learning
• Cross-Domain Few-Shot Learning by Representation Fusion
• Adversarial Feature Hallucination Networks for Few-Shot Learning
• Node Classification on Graphs with Few-Shot Novel Labels via Meta Transformed Network Embedding
• A Baseline for Few-Shot Image Classification
• Explanation-Guided Training for Cross-Domain Few-Shot Classification
• Enhancing Few-shot Image Classification with Cosine Transformer
• Towards Fair Knowledge Transfer for Imbalanced Domain Adaptation
• RelationNet2: Deep Comparison Columns for Few-Shot Learning
• Learning to Focus: Cascaded Feature Matching Network for Few-shot Image Recognition
• Self-training for Few-shot Transfer Across Extreme Task Differences
• Towards Cross-Granularity Few-Shot Learning: Coarse-to-Fine Pseudo-Labeling with Visual-Semantic Meta-Embedding
• Relational Generalized Few-Shot Learning
• Prototype Rectification for Few-Shot Learning
• Meta-Learning for Short Utterance Speaker Recognition with Imbalance Length Pairs
• SAC: Accelerating and Structuring Self-Attention via Sparse Adaptive Connection
• Learning to learn via Self-Critique
• Are Few-Shot Learning Benchmarks too Simple ? Solving them without Task Supervision at Test-Time
• Label Propagation with Augmented Anchors: A Simple Semi-Supervised Learning baseline for Unsupervised Domain Adaptation
• Region Comparison Network for Interpretable Few-shot Image Classification
• Dynamic Distillation Network for Cross-Domain Few-Shot Recognition with Unlabeled Data
• L2AE-D: Learning to Aggregate Embeddings for Few-shot Learning with Meta-level Dropout
• Few-Shot Learning as Domain Adaptation: Algorithm and Analysis
• MetaFun: Meta-Learning with Iterative Functional Updates
• Learning to Affiliate: Mutual Centralized Learning for Few-shot Classification
• Graph-based Interpolation of Feature Vectors for Accurate Few-Shot Classification
• Learning from the Past: Continual Meta-Learning via Bayesian Graph Modeling
• Online Structured Meta-learning
• Learning from Adversarial Features for Few-Shot Classification
• Prototype Completion with Primitive Knowledge for Few-Shot Learning
• TransMatch: A Transfer-Learning Scheme for Semi-Supervised Few-Shot Learning
• Learning from Multiple Cities: A Meta-Learning Approach for Spatial-Temporal Prediction
• Cross-Domain Few-Shot Classification via Adversarial Task Augmentation
• XtarNet: Learning to Extract Task-Adaptive Representation for Incremental Few-Shot Learning
• ECKPN: Explicit Class Knowledge Propagation Network for Transductive Few-shot Learning
• Meta-Learning with Adaptive Hyperparameters
• Generalized Zero and Few-Shot Transfer for Facial Forgery Detection
• Attentive Graph Neural Networks for Few-Shot Learning
• Self-Prediction for Joint Instance and Semantic Segmentation of Point Clouds
• Temporal Transductive Inference for Few-Shot Video Object Segmentation
• Semi-Supervised Few-Shot Learning with Prototypical Random Walks
• Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time
• Why Propagate Alone? Parallel Use of Labels and Features on Graphs
• Adaptive Task Sampling for Meta-Learning
• Continual Local Replacement for Few-shot Learning
• Hierarchical Representation based Query-Specific Prototypical Network for Few-Shot Image Classification
• Revisiting Metric Learning for Few-Shot Image Classification
• Multi-Level Correlation Network For Few-Shot Image Classification
• AdarGCN: Adaptive Aggregation GCN for Few-Shot Learning
• Ranking Distance Calibration for Cross-Domain Few-Shot Learning
• Unsupervised Meta-Learning through Latent-Space Interpolation in Generative Models
• CAFENet: Class-Agnostic Few-Shot Edge Detection Network
• Memory-Based Neighbourhood Embedding for Visual Recognition
• Bidirectional RNN-based Few Shot Learning for 3D Medical Image Segmentation
• Complementing Representation Deficiency in Few-shot Image Classification: A Meta-Learning Approach
• How to trust unlabeled data? Instance Credibility Inference for Few-Shot Learning
• Temporal Alignment Prediction for Few-Shot Video Classification
• Shot in the Dark: Few-Shot Learning with No Base-Class Labels
• A Strong Baseline for Semi-Supervised Incremental Few-Shot Learning
• Few-shot learning via tensor hallucination
• Modular Adaptation for Cross-Domain Few-Shot Learning
• Few-shot 3D Point Cloud Semantic Segmentation
• Graph convolutional networks for learning with few clean and many noisy labels
• Task-Adaptive Clustering for Semi-Supervised Few-Shot Classification
• Trainable Class Prototypes for Few-Shot Learning
• Meta Two-Sample Testing: Learning Kernels for Testing with Limited Data
• Few-shot Learning with Global Relatedness Decoupled-Distillation
• Memory-Augmented Relation Network for Few-Shot Learning
• Will Multi-modal Data Improves Few-shot Learning?
• Progressive Cluster Purification for Transductive Few-shot Learning
• Extended Few-Shot Learning: Exploiting Existing Resources for Novel Tasks
• Large-Scale Meta-Learning with Continual Trajectory Shifting
• Mutual-Information Based Few-Shot Classification
• Learning by Examples Based on Multi-level Optimization
• Meta-Learning with Task-Adaptive Loss Function for Few-Shot Learning
• Transductive Learning for Textual Few-Shot Classification in API-based Embedding Models
• Learn from Anywhere: Rethinking Generalized Zero-Shot Learning with Limited Supervision
• Few-shot Learning with LSSVM Base Learner and Transductive Modules
• MDFM: Multi-Decision Fusing Model for Few-Shot Learning
• Diversity Transfer Network for Few-Shot Learning
• A Transductive Maximum Margin Classifier for Few-Shot Learning
• Shoestring: Graph-Based Semi-Supervised Learning with Severely Limited Labeled Data
• Episode Adaptive Embedding Networks for Few-shot Learning
• Dataset Bias in Few-shot Image Recognition
• Class Interference Regularization