Infinite Mixture Prototypes for Few-Shot Learning

Abstract

We propose infinite mixture prototypes to adaptively represent both simple and complex data distributions for few-shot learning. Our infinite mixture prototypes represent each class by a set of clusters, unlike existing prototypical methods that represent each class by a single cluster. By inferring the number of clusters, infinite mixture prototypes interpolate between nearest neighbor and prototypical representations, which improves accuracy and robustness in the few-shot regime. We show the importance of adaptive capacity for capturing complex data distributions such as alphabets, with 25% absolute accuracy improvements over prototypical networks, while still maintaining or improving accuracy on the standard Omniglot and mini-ImageNet benchmarks. In clustering labeled and unlabeled data by the same clustering rule, infinite mixture prototypes achieves state-of-the-art semi-supervised accuracy. As a further capability, we show that infinite mixture prototypes can perform purely unsupervised clustering, unlike existing prototypical methods.

Cited by in corpus (44)

• BSNet: Bi-Similarity Network for Few-shot Fine-grained Image Classification
• Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need?
• Self-supervised Knowledge Distillation for Few-shot Learning
• Hypercorrelation Squeeze for Few-Shot Segmentation
• Many-Class Few-Shot Learning on Multi-Granularity Class Hierarchy
• Prototype Guided Federated Learning of Visual Feature Representations
• DPGN: Distribution Propagation Graph Network for Few-shot Learning
• Cross-Domain Few-Shot Learning by Representation Fusion
• Prototype Refinement Network for Few-Shot Segmentation
• Learning to Learn Variational Semantic Memory
• Wandering Within a World: Online Contextualized Few-Shot Learning
• On Episodes, Prototypical Networks, and Few-shot Learning
• Multi-scale Adaptive Task Attention Network for Few-Shot Learning
• Prototype Rectification for Few-Shot Learning
• A Closer Look at Prototype Classifier for Few-shot Image Classification
• Meta-Learning across Meta-Tasks for Few-Shot Learning
• A Theoretical Analysis of the Number of Shots in Few-Shot Learning
• Mutual exclusivity as a challenge for deep neural networks
• Distance-wise Prototypical Graph Neural Network in Node Imbalance Classification
• Contrastive Prototype Learning with Augmented Embeddings for Few-Shot Learning
• Task-similarity Aware Meta-learning through Nonparametric Kernel Regression
• Dynamic Memory Induction Networks for Few-Shot Text Classification
• Few-shot tweet detection in emerging disaster events
• On the Importance of Firth Bias Reduction in Few-Shot Classification
• Concept Learners for Few-Shot Learning
• Prototypical Networks for Multi-Label Learning
• Hierarchical Representation based Query-Specific Prototypical Network for Few-Shot Image Classification
• Function Contrastive Learning of Transferable Meta-Representations
• AdarGCN: Adaptive Aggregation GCN for Few-Shot Learning
• Few-Shot Learning with Intra-Class Knowledge Transfer
• Asymmetric Distribution Measure for Few-shot Learning
• Introducing the structural bases of typicality effects in deep learning
• Multi-Label Few-Shot Learning for Aspect Category Detection
• Interpretable Time-series Classification on Few-shot Samples
• Global Semantic Description of Objects based on Prototype Theory
• Variable-Shot Adaptation for Online Meta-Learning
• Fuzzy Simplicial Networks: A Topology-Inspired Model to Improve Task Generalization in Few-shot Learning
• Enabling the Network to Surf the Internet
• Task-Adaptive Clustering for Semi-Supervised Few-Shot Classification
• Online Unsupervised Learning of Visual Representations and Categories
• Meta-Learning with Adjoint Methods
• One-shot Learning with Absolute Generalization
• Probabilistic task modelling for meta-learning
• Temporal Action Segmentation with High-level Complex Activity Labels