Matching Networks for One Shot Learning

Abstract

Learning from a few examples remains a key challenge in machine learning. Despite recent advances in important domains such as vision and language, the standard supervised deep learning paradigm does not offer a satisfactory solution for learning new concepts rapidly from little data. In this work, we employ ideas from metric learning based on deep neural features and from recent advances that augment neural networks with external memories. Our framework learns a network that maps a small labelled support set and an unlabelled example to its label, obviating the need for fine-tuning to adapt to new class types. We then define one-shot learning problems on vision (using Omniglot, ImageNet) and language tasks. Our algorithm improves one-shot accuracy on ImageNet from 87.6% to 93.2% and from 88.0% to 93.8% on Omniglot compared to competing approaches. We also demonstrate the usefulness of the same model on language modeling by introducing a one-shot task on the Penn Treebank.

References in corpus (3)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Sequence to Sequence Learning with Neural Networks
• Recurrent Neural Network Regularization

Cited by in corpus (274)

• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• GOT-10k: A Large High-Diversity Benchmark for Generic Object Tracking in the Wild
• Energy-based Generative Adversarial Network
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Deep Reinforcement Learning: An Overview
• Few-Shot Learning with Graph Neural Networks
• Federated Meta-Learning with Fast Convergence and Efficient Communication
• SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning
• One-Shot Imitation Learning
• BSNet: Bi-Similarity Network for Few-shot Fine-grained Image Classification
• Few-shot classification in Named Entity Recognition Task
• Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination
• Recasting Gradient-Based Meta-Learning as Hierarchical Bayes
• Bayesian Model-Agnostic Meta-Learning
• Continuous Adaptation via Meta-Learning in Nonstationary and Competitive Environments
• Towards a Neural Statistician
• Making Pre-trained Language Models Better Few-shot Learners
• Deep Learning in Mobile and Wireless Networking: A Survey
• Robust High-dimensional Memory-augmented Neural Networks
• Learning to Generalize: Meta-Learning for Domain Generalization
• MetaKG: Meta-learning on Knowledge Graph for Cold-start Recommendation
• Mechanical Search: Multi-Step Retrieval of a Target Object Occluded by Clutter
• Bilevel Programming for Hyperparameter Optimization and Meta-Learning
• Deep Meta-Learning: Learning to Learn in the Concept Space
• Few-Shot Segmentation Propagation with Guided Networks
• Few-Shot Deep Adversarial Learning for Video-based Person Re-identification
• Learning to Learn: Meta-Critic Networks for Sample Efficient Learning
• Towards a universal neural network encoder for time series
• Neural Models for Information Retrieval
• xView: Objects in Context in Overhead Imagery
• Generative Adversarial Residual Pairwise Networks for One Shot Learning
• Gotta Learn Fast: A New Benchmark for Generalization in RL
• Attentive Recurrent Comparators
• Dynamic Few-Shot Visual Learning without Forgetting
• Self-supervised Knowledge Distillation for Few-shot Learning
• Decoupled Novel Object Captioner
• Low-Shot Learning from Imaginary Data
• Active One-shot Learning
• Few-Shot Learning Through an Information Retrieval Lens
• Recurrent Neural Networks for Semantic Instance Segmentation
• Learning similarity measures from data
• Cognitive Psychology for Deep Neural Networks: A Shape Bias Case Study
• Zero-Shot Visual Question Answering
• MaskConnect: Connectivity Learning by Gradient Descent
• Label Efficient Learning of Transferable Representations across Domains and Tasks
• Learning Algorithms for Active Learning
• Conditional Neural Processes
• Zero-Shot Learning via Class-Conditioned Deep Generative Models
• Auto-Ensemble: An Adaptive Learning Rate Scheduling based Deep Learning Model Ensembling
• Self-Augmentation: Generalizing Deep Networks to Unseen Classes for Few-Shot Learning
• CANet: Class-Agnostic Segmentation Networks with Iterative Refinement and Attentive Few-Shot Learning
• FewRel: A Large-Scale Supervised Few-Shot Relation Classification Dataset with State-of-the-Art Evaluation
• Reward Shaping via Meta-Learning
• Comparison of Maximum Likelihood and GAN-based training of Real NVPs
• Leveraging Siamese Networks for One-Shot Intrusion Detection Model
• Efficient Video Object Segmentation via Network Modulation
• Few-shot Autoregressive Density Estimation: Towards Learning to Learn Distributions
• Neural Data Augmentation via Example Extrapolation
• CleanNet: Transfer Learning for Scalable Image Classifier Training with Label Noise
• Memory Matching Networks for One-Shot Image Recognition
• Few-Shot Goal Inference for Visuomotor Learning and Planning
• Meta-GNN: On Few-shot Node Classification in Graph Meta-learning
• Deep Weakly-supervised Anomaly Detection
• Prior-Knowledge and Attention-based Meta-Learning for Few-Shot Learning
• Deep Triplet Ranking Networks for One-Shot Recognition
• Been There, Done That: Meta-Learning with Episodic Recall
• Leveraging the Feature Distribution in Transfer-based Few-Shot Learning
• TaskNorm: Rethinking Batch Normalization for Meta-Learning
• Low-shot Learning via Covariance-Preserving Adversarial Augmentation Networks
• A Deep One-Shot Network for Query-based Logo Retrieval
• Edge-labeling Graph Neural Network for Few-shot Learning
• Meta-Learned Confidence for Few-shot Learning
• The Variational Homoencoder: Learning to learn high capacity generative models from few examples
• Meta-Learning for Low-Resource Neural Machine Translation
• Auto-Encoding Scene Graphs for Image Captioning
• Feature Transfer Learning for Deep Face Recognition with Under-Represented Data
• Meta-Learning by the Baldwin Effect
• Auto-Meta: Automated Gradient Based Meta Learner Search
• Diverse Few-Shot Text Classification with Multiple Metrics
• Meta-DETR: Image-Level Few-Shot Object Detection with Inter-Class Correlation Exploitation
• Probabilistic Model-Agnostic Meta-Learning
• Large Margin Few-Shot Learning
• Learning to Warm-Start Bayesian Hyperparameter Optimization
• Weakly-supervised Compositional FeatureAggregation for Few-shot Recognition
• General stochastic separation theorems with optimal bounds
• CTU Depth Decision Algorithms for HEVC: A Survey
• Metalearning with Hebbian Fast Weights
• Fast Parametric Learning with Activation Memorization
• Backdoor Attacks on Federated Meta-Learning
• Deep Nets: What have they ever done for Vision?
• Few-Shot Image Recognition by Predicting Parameters from Activations
• Multi-class Classification without Multi-class Labels
• One-Shot Relational Learning for Knowledge Graphs
• Few-Shot Video Classification via Temporal Alignment
• Learning models for visual 3D localization with implicit mapping
• Max-margin Class Imbalanced Learning with Gaussian Affinity
• Transfer Learning for Estimating Causal Effects using Neural Networks
• Meta-Learning for Low-resource Natural Language Generation in Task-oriented Dialogue Systems
• Generative One-Shot Learning (GOL): A Semi-Parametric Approach to One-Shot Learning in Autonomous Vision
• Neural Task Graphs: Generalizing to Unseen Tasks from a Single Video Demonstration
• Few-Shot Class-Incremental Learning
• Generalization Bounds For Meta-Learning: An Information-Theoretic Analysis
• Associative Compression Networks for Representation Learning
• Lightweight Speaker Verification for Online Identification of New Speakers with Short Segments
• Dynamic backdoor attacks against federated learning
• ACE: Adapting to Changing Environments for Semantic Segmentation
• Few-Shot Scene Adaptive Crowd Counting Using Meta-Learning
• Toward Multimodal Model-Agnostic Meta-Learning
• Large Margin Deep Networks for Classification
• Online Adaptation through Meta-Learning for Stereo Depth Estimation
• Lung Cancer Screening Using Adaptive Memory-Augmented Recurrent Networks
• Meta-learning autoencoders for few-shot prediction
• TAFE-Net: Task-Aware Feature Embeddings for Low Shot Learning
• Reconciling meta-learning and continual learning with online mixtures of tasks
• Towards Understanding Generalization in Gradient-Based Meta-Learning
• $k$-Nearest Neighbor Augmented Neural Networks for Text Classification
• LCNN: Lookup-based Convolutional Neural Network
• The Kanerva Machine: A Generative Distributed Memory
• Graph Meta Learning via Local Subgraphs
• When Low Resource NLP Meets Unsupervised Language Model: Meta-pretraining Then Meta-learning for Few-shot Text Classification
• One-Shot Learning using Mixture of Variational Autoencoders: a Generalization Learning approach
• Improving Generalization via Scalable Neighborhood Component Analysis
• Fast Adaptation in Generative Models with Generative Matching Networks
• Region Comparison Network for Interpretable Few-shot Image Classification
• Uncertainty in Multitask Transfer Learning
• Dynamic Distillation Network for Cross-Domain Few-Shot Recognition with Unlabeled Data
• Blessing of dimensionality at the edge
• Variational Prototyping-Encoder: One-Shot Learning with Prototypical Images
• It Takes Two to Tango: Mixup for Deep Metric Learning
• Zero-Shot Learning from scratch (ZFS): leveraging local compositional representations
• Task-Aware Feature Generation for Zero-Shot Compositional Learning
• Meta-Amortized Variational Inference and Learning
• Deep Co-attention based Comparators For Relative Representation Learning in Person Re-identification
• A Computational Model of Commonsense Moral Decision Making
• ARCADe: A Rapid Continual Anomaly Detector
• Meta-learnt priors slow down catastrophic forgetting in neural networks
• AHA! an 'Artificial Hippocampal Algorithm' for Episodic Machine Learning
• MetaPred: Meta-Learning for Clinical Risk Prediction with Limited Patient Electronic Health Records
• Is Support Set Diversity Necessary for Meta-Learning?
• Low-Shot Learning with Imprinted Weights
• Learning to Learn with Feedback and Local Plasticity
• Prototypical Clustering Networks for Dermatological Disease Diagnosis
• Few-shot Slot Tagging with Collapsed Dependency Transfer and Label-enhanced Task-adaptive Projection Network
• Learning from scarce information: using synthetic data to classify Roman fine ware pottery
• Towards Effective Deep Embedding for Zero-Shot Learning
• Differentiable Meta-learning Model for Few-shot Semantic Segmentation
• GPU-based Self-Organizing Maps for Post-Labeled Few-Shot Unsupervised Learning
• Learning From Less Data: Diversified Subset Selection and Active Learning in Image Classification Tasks
• Few-shot Network Anomaly Detection via Cross-network Meta-learning
• Learning from similarity and information extraction from structured documents
• Active learning for interactive satellite image change detection
• Weakly-supervised Object Localization for Few-shot Learning and Fine-grained Few-shot Learning
• AgileNet: Lightweight Dictionary-based Few-shot Learning
• Generalized Adaptation for Few-Shot Learning
• Few-Shot Knowledge Graph Completion
• Sharkzor: Interactive Deep Learning for Image Triage, Sort and Summary
• Few-Shot Open-Set Recognition using Meta-Learning
• Generating new concepts with hybrid neuro-symbolic models
• CURI: A Benchmark for Productive Concept Learning Under Uncertainty
• Chameleon: Learning Model Initializations Across Tasks With Different Schemas
• Meta Learning for Few-Shot One-class Classification
• Machine learning with limited data
• Non-Gaussian Gaussian Processes for Few-Shot Regression
• Zero and Few Shot Learning with Semantic Feature Synthesis and Competitive Learning
• Cross-Entropy Adversarial View Adaptation for Person Re-identification
• Active Object Manipulation Facilitates Visual Object Learning: An Egocentric Vision Study
• Making Good on LSTMs' Unfulfilled Promise
• Graph Few-shot Learning via Knowledge Transfer
• Boosting Supervision with Self-Supervision for Few-shot Learning
• Building Machines that Learn and Think for Themselves: Commentary on Lake et al., Behavioral and Brain Sciences, 2017
• An empirical study of pretrained representations for few-shot classification
• Multi-Source Domain Adaptation with Mixture of Experts
• Learning to Support: Exploiting Structure Information in Support Sets for One-Shot Learning
• Deep Bilevel Learning
• A Study of Few-Shot Audio Classification
• Information-Theoretic Analysis of Epistemic Uncertainty in Bayesian Meta-learning
• Semi-Supervised and Active Few-Shot Learning with Prototypical Networks
• Personalized acute stress classification from physiological signals with neural processes
• MetaSelector: Meta-Learning for Recommendation with User-Level Adaptive Model Selection
• Flexible Compositional Learning of Structured Visual Concepts
• Few-Shot Semantic Segmentation Augmented with Image-Level Weak Annotations
• SML: Semantic Meta-learning for Few-shot Semantic Segmentation
• HetMAML: Task-Heterogeneous Model-Agnostic Meta-Learning for Few-Shot Learning Across Modalities
• Featurized Density Ratio Estimation
• Faster Optimization-Based Meta-Learning Adaptation Phase
• What Can Knowledge Bring to Machine Learning? -- A Survey of Low-shot Learning for Structured Data
• Physics-aware Spatiotemporal Modules with Auxiliary Tasks for Meta-Learning
• Stress Testing of Meta-learning Approaches for Few-shot Learning
• Few-Shot NLU with Vector Projection Distance and Abstract Triangular CRF
• MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
• Single-View 3D Object Reconstruction from Shape Priors in Memory
• Long-term Cross Adversarial Training: A Robust Meta-learning Method for Few-shot Classification Tasks
• Adversarial Monte Carlo Meta-Learning of Optimal Prediction Procedures
• Learning to Customize Model Structures for Few-shot Dialogue Generation Tasks
• Learning Continually from Low-shot Data Stream
• HalalNet: A Deep Neural Network that Classifies the Halalness Slaughtered Chicken from their Images
• SimulLR: Simultaneous Lip Reading Transducer with Attention-Guided Adaptive Memory
• Dynamic Key-Value Memory Networks for Knowledge Tracing
• ST-MAML: A Stochastic-Task based Method for Task-Heterogeneous Meta-Learning
• Data-Efficient Graph Embedding Learning for PCB Component Detection
• Adaptive Prototypical Networks with Label Words and Joint Representation Learning for Few-Shot Relation Classification
• Contextual HyperNetworks for Novel Feature Adaptation
• Few-shot Intent Classification and Slot Filling with Retrieved Examples
• Few-shot Continual Learning: a Brain-inspired Approach
• How many winning tickets are there in one DNN?
• Learning-to-Learn Personalised Human Activity Recognition Models
• Few-Shot Object Recognition from Machine-Labeled Web Images
• Few-shot Object Detection with Self-adaptive Attention Network for Remote Sensing Images
• Learning to generate classifiers
• Asymmetric Distribution Measure for Few-shot Learning
• Towards Collaborative Intelligence Friendly Architectures for Deep Learning
• AuxBlocks: Defense Adversarial Example via Auxiliary Blocks
• Meta-Learning via Feature-Label Memory Network
• Minimally-Supervised Structure-Rich Text Categorization via Learning on Text-Rich Networks
• Meta-learning One-class Classifiers with Eigenvalue Solvers for Supervised Anomaly Detection
• Federated Few-Shot Learning with Adversarial Learning
• Reliable Label Bootstrapping for Semi-Supervised Learning
• Generalized Coarse-to-Fine Visual Recognition with Progressive Training
• Representation based and Attention augmented Meta learning
• Marrying up Regular Expressions with Neural Networks: A Case Study for Spoken Language Understanding
• Towards Robust Pattern Recognition: A Review
• Retrosynthetic reaction prediction using neural sequence-to-sequence models
• Proximal Mapping for Deep Regularization
• Tackling Early Sparse Gradients in Softmax Activation Using Leaky Squared Euclidean Distance
• Few-Shot Event Detection with Prototypical Amortized Conditional Random Field
• Learning Class-level Prototypes for Few-shot Learning
• Few-Sample Named Entity Recognition for Security Vulnerability Reports by Fine-Tuning Pre-Trained Language Models
• Neural Conditional Gradients
• Uncertainty-Aware Few-Shot Image Classification
• Will Multi-modal Data Improves Few-shot Learning?
• Modular meta-learning in abstract graph networks for combinatorial generalization
• Low-Shot Learning from Imaginary 3D Model
• Continual Learning Augmented Investment Decisions
• f-VAEGAN-D2: A Feature Generating Framework for Any-Shot Learning
• A Relational Model for One-Shot Classification
• COFGA: Classification Of Fine-Grained Features In Aerial Images
• One-Shot Image-to-Image Translation via Part-Global Learning with a Multi-adversarial Framework
• Categorizing Items with Short and Noisy Descriptions using Ensembled Transferred Embeddings
• End-to-end Generative Zero-shot Learning via Few-shot Learning
• One-shot domain adaptation for semantic face editing of real world images using StyleALAE
• Constructing Multiple Tasks for Augmentation: Improving Neural Image Classification With K-means Features
• Set2Model Networks: Learning Discriminatively To Learn Generative Models
• Learning adaptively from the unknown for few-example video person re-ID
• Stochastic Whitening Batch Normalization
• Edge-Labeling based Directed Gated Graph Network for Few-shot Learning
• Learning to Generalize to Unseen Tasks with Bilevel Optimization
• One-Shot Induction of Generalized Logical Concepts via Human Guidance
• Meta learning to classify intent and slot labels with noisy few shot examples
• Scaling Laws for the Few-Shot Adaptation of Pre-trained Image Classifiers
• Are You Sure You Want To Do That? Classification with Verification
• Few-Shot Adaptation for Multimedia Semantic Indexing
• Reinforced Attention for Few-Shot Learning and Beyond
• Multi-accent Speech Separation with One Shot Learning
• Robust Task Clustering for Deep Many-Task Learning
• On sensitivity of meta-learning to support data
• One-Shot Image Classification by Learning to Restore Prototypes
• Strategies for Language Identification in Code-Mixed Low Resource Languages
• Forward Amortized Inference for Likelihood-Free Variational Marginalization
• Fine-Grained Visual Categorization using Meta-Learning Optimization with Sample Selection of Auxiliary Data
• Power Normalizing Second-order Similarity Network for Few-shot Learning
• mil-benchmarks: Standardized Evaluation of Deep Multiple-Instance Learning Techniques
• Deep Episodic Value Iteration for Model-based Meta-Reinforcement Learning
• HMRL: Hyper-Meta Learning for Sparse Reward Reinforcement Learning Problem
• Generating meta-learning tasks to evolve parametric loss for classification learning
• Meta-Learning for Koopman Spectral Analysis with Short Time-series
• Mutual-Information Based Few-Shot Classification
• Meta-learning for downstream aware and agnostic pretraining
• User Intent Classification using Memory Networks: A Comparative Analysis for a Limited Data Scenario
• One-Shot Learning for Language Modelling
• Leveraging Table Content for Zero-shot Text-to-SQL with Meta-Learning
• Few-shot Learning via Dependency Maximization and Instance Discriminant Analysis
• Fair Meta-Learning For Few-Shot Classification
• Learning to Learn Image Classifiers with Visual Analogy
• InsectUp: Crowdsourcing Insect Observations to Assess Demographic Shifts and Improve Classification