Meta-SGD: Learning to Learn Quickly for Few-Shot Learning

Abstract

Few-shot learning is challenging for learning algorithms that learn each task in isolation and from scratch. In contrast, meta-learning learns from many related tasks a meta-learner that can learn a new task more accurately and faster with fewer examples, where the choice of meta-learners is crucial. In this paper, we develop Meta-SGD, an SGD-like, easily trainable meta-learner that can initialize and adapt any differentiable learner in just one step, on both supervised learning and reinforcement learning. Compared to the popular meta-learner LSTM, Meta-SGD is conceptually simpler, easier to implement, and can be learned more efficiently. Compared to the latest meta-learner MAML, Meta-SGD has a much higher capacity by learning to learn not just the learner initialization, but also the learner update direction and learning rate, all in a single meta-learning process. Meta-SGD shows highly competitive performance for few-shot learning on regression, classification, and reinforcement learning.

reinforcement learning included, 20-way classification on MiniImagenet included

References in corpus (8)

• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Theoretical Models of Learning to Learn
• CNN Features off-the-shelf: an Astounding Baseline for Recognition
• RL${}^2$: Fast Reinforcement Learning via Slow Reinforcement Learning
• Learning to reinforcement learn
• Learning to Optimize
• Learning to Learn: Meta-Critic Networks for Sample Efficient Learning

Cited by in corpus (221)

• Personalized Federated Learning: A Meta-Learning Approach
• Federated Meta-Learning with Fast Convergence and Efficient Communication
• SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning
• Meta-Learning with Latent Embedding Optimization
• Free Lunch for Few-shot Learning: Distribution Calibration
• Meta-Reinforcement Learning of Structured Exploration Strategies
• Piecewise classifier mappings: Learning fine-grained learners for novel categories with few examples
• An Ensemble of Epoch-wise Empirical Bayes for Few-shot Learning
• Deep Meta-Learning: Learning to Learn in the Concept Space
• Finding Task-Relevant Features for Few-Shot Learning by Category Traversal
• Self-Supervised Generalisation with Meta Auxiliary Learning
• Learning to Optimize: A Primer and A Benchmark
• How to train your MAML
• Empirical Bayes Transductive Meta-Learning with Synthetic Gradients
• Feature-Critic Networks for Heterogeneous Domain Generalization
• Self-supervised Knowledge Distillation for Few-shot Learning
• From Learning to Meta-Learning: Reduced Training Overhead and Complexity for Communication Systems
• Test-Time Training with Self-Supervision for Generalization under Distribution Shifts
• Meta-Learning with Warped Gradient Descent
• One-Shot Instance Segmentation
• Transfer Learning via Contextual Invariants for One-to-Many Cross-Domain Recommendation
• Dual Adaptive Representation Alignment for Cross-domain Few-shot Learning
• The Internet of Federated Things (IoFT): A Vision for the Future and In-depth Survey of Data-driven Approaches for Federated Learning
• Multi-level Second-order Few-shot Learning
• Small Sample Learning in Big Data Era
• SGMNet: Scene Graph Matching Network for Few-Shot Remote Sensing Scene Classification
• Alpha MAML: Adaptive Model-Agnostic Meta-Learning
• MeLU: Meta-Learned User Preference Estimator for Cold-Start Recommendation
• Federated Learning with Unbiased Gradient Aggregation and Controllable Meta Updating
• MetaAvatar: Learning Animatable Clothed Human Models from Few Depth Images
• Rectifying the Shortcut Learning of Background for Few-Shot Learning
• DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning
• Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs
• Learning from Few Samples: A Survey
• learn2learn: A Library for Meta-Learning Research
• A Comprehensive Overview and Survey of Recent Advances in Meta-Learning
• Transductive Episodic-Wise Adaptive Metric for Few-Shot Learning
• La-MAML: Look-ahead Meta Learning for Continual Learning
• Personalized Federated Learning using Hypernetworks
• Provable Guarantees for Gradient-Based Meta-Learning
• Prior-Knowledge and Attention-based Meta-Learning for Few-Shot Learning
• Meta-Transfer Learning for Few-Shot Learning
• Theoretical Convergence of Multi-Step Model-Agnostic Meta-Learning
• Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition from a Domain Adaptation Perspective
• Controllable Semantic Parsing via Retrieval Augmentation
• Defining Benchmarks for Continual Few-Shot Learning
• Re-Scoring Using Image-Language Similarity for Few-Shot Object Detection
• Meta-Learning Representations for Continual Learning
• Task-Robust Model-Agnostic Meta-Learning
• Adversarial Feature Hallucination Networks for Few-Shot Learning
• Probabilistic Model-Agnostic Meta-Learning
• Image Deformation Meta-Networks for One-Shot Learning
• Meta-Learning Symmetries by Reparameterization
• Large Margin Few-Shot Learning
• Automated Relational Meta-learning
• Adaptive Gradient-Based Meta-Learning Methods
• MobilePhys: Personalized Mobile Camera-Based Contactless Physiological Sensing
• Learning to Balance: Bayesian Meta-Learning for Imbalanced and Out-of-distribution Tasks
• RelationNet2: Deep Comparison Columns for Few-Shot Learning
• Bayesian Active Meta-Learning for Few Pilot Demodulation and Equalization
• Learning to Learn Variational Semantic Memory
• One-Shot Relational Learning for Knowledge Graphs
• Adaptive Deep Kernel Learning
• DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection
• Wandering Within a World: Online Contextualized Few-Shot Learning
• Revisiting Meta-Learning as Supervised Learning
• Towards Cross-Granularity Few-Shot Learning: Coarse-to-Fine Pseudo-Labeling with Visual-Semantic Meta-Embedding
• Learning Dynamic Alignment via Meta-filter for Few-shot Learning
• When MAML Can Adapt Fast and How to Assist When It Cannot
• FLIP: A flexible initializer for arbitrarily-sized parametrized quantum circuits
• Relational Generalized Few-Shot Learning
• Language-Inspired Relation Transfer for Few-shot Class-Incremental Learning
• Prototype Rectification for Few-Shot Learning
• Operator Learning with Neural Fields: Tackling PDEs on General Geometries
• Learning Causal Models Online
• MT3: Meta Test-Time Training for Self-Supervised Test-Time Adaption
• Learning Adaptive Classifiers Synthesis for Generalized Few-Shot Learning
• MetaTroll: Few-shot Detection of State-Sponsored Trolls with Transformer Adapters
• Few-shot Learning for Time-series Forecasting
• On the Importance of Attention in Meta-Learning for Few-Shot Text Classification
• Toward Multimodal Model-Agnostic Meta-Learning
• Few-shot Classification via Adaptive Attention
• A Generic First-Order Algorithmic Framework for Bi-Level Programming Beyond Lower-Level Singleton
• Deep Metric Transfer for Label Propagation with Limited Annotated Data
• Meta-Learning across Meta-Tasks for Few-Shot Learning
• Fixed-MAML for Few Shot Classification in Multilingual Speech Emotion Recognition
• Meta-learning autoencoders for few-shot prediction
• Improving Generalization in Meta-learning via Task Augmentation
• Learning to learn via Self-Critique
• MetaPhys: Few-Shot Adaptation for Non-Contact Physiological Measurement
• Personalized Federated Learning with Gaussian Processes
• Meta-Learning with Fewer Tasks through Task Interpolation
• Scalable Bayesian Meta-Learning through Generalized Implicit Gradients
• DiCE: The Infinitely Differentiable Monte-Carlo Estimator
• Improving Generalization via Scalable Neighborhood Component Analysis
• Region Comparison Network for Interpretable Few-shot Image Classification
• Context Meta-Reinforcement Learning via Neuromodulation
• StyleMeUp: Towards Style-Agnostic Sketch-Based Image Retrieval
• Dual-Awareness Attention for Few-Shot Object Detection
• Stateless Neural Meta-Learning using Second-Order Gradients
• Expert Training: Task Hardness Aware Meta-Learning for Few-Shot Classification
• Meta Reinforcement Learning with Task Embedding and Shared Policy
• MetaFun: Meta-Learning with Iterative Functional Updates
• An Investigation of Few-Shot Learning in Spoken Term Classification
• Instance Credibility Inference for Few-Shot Learning
• Weighted Meta-Learning
• Few-Shot Learning as Domain Adaptation: Algorithm and Analysis
• Modular Meta-Learning with Shrinkage
• Online Structured Meta-learning
• Learning from the Past: Continual Meta-Learning via Bayesian Graph Modeling
• ARCADe: A Rapid Continual Anomaly Detector
• Memory-Based Optimization Methods for Model-Agnostic Meta-Learning and Personalized Federated Learning
• Storchastic: A Framework for General Stochastic Automatic Differentiation
• Meta-learning algorithms for Few-Shot Computer Vision
• LaSO: Label-Set Operations networks for multi-label few-shot learning
• Meta-Learning PAC-Bayes Priors in Model Averaging
• Structured Prediction for Conditional Meta-Learning
• Generalized Adaptation for Few-Shot Learning
• Regularizing Meta-Learning via Gradient Dropout
• Formulating Camera-Adaptive Color Constancy as a Few-shot Meta-Learning Problem
• Adaptive-Step Graph Meta-Learner for Few-Shot Graph Classification
• Multi-objective Neural Architecture Search via Non-stationary Policy Gradient
• Few-Shot Electronic Health Record Coding through Graph Contrastive Learning
• Meta-Learning Initializations for Image Segmentation
• Online Meta-Learning for Multi-Source and Semi-Supervised Domain Adaptation
• Few-Shot Knowledge Graph Completion
• EvoGrad: Efficient Gradient-Based Meta-Learning and Hyperparameter Optimization
• Adaptive Task Sampling for Meta-Learning
• Embedding Adaptation is Still Needed for Few-Shot Learning
• How Fine-Tuning Allows for Effective Meta-Learning
• Guarantees for Tuning the Step Size using a Learning-to-Learn Approach
• Rectified Meta-Learning from Noisy Labels for Robust Image-based Plant Disease Diagnosis
• Adaptive Cross-Modal Few-Shot Learning
• Cross-Modulation Networks for Few-Shot Learning
• Function Contrastive Learning of Transferable Meta-Representations
• Semi-Supervised and Active Few-Shot Learning with Prototypical Networks
• MetaSelector: Meta-Learning for Recommendation with User-Level Adaptive Model Selection
• On the Importance of Firth Bias Reduction in Few-Shot Classification
• Generalization of Model-Agnostic Meta-Learning Algorithms: Recurring and Unseen Tasks
• Real-Time Object Tracking via Meta-Learning: Efficient Model Adaptation and One-Shot Channel Pruning
• Unsupervised Meta-Testing with Conditional Neural Processes for Hybrid Meta-Reinforcement Learning
• Meta-learning with negative learning rates
• Meta Dropout: Learning to Perturb Features for Generalization
• Evaluating the progress of Deep Reinforcement Learning in the real world: aligning domain-agnostic and domain-specific research
• A Channel Coding Benchmark for Meta-Learning
• Faster Optimization-Based Meta-Learning Adaptation Phase
• MPLP: Learning a Message Passing Learning Protocol
• Automatic Validation of Textual Attribute Values in E-commerce Catalog by Learning with Limited Labeled Data
• Joint Image-Instance Spatial-Temporal Attention for Few-shot Action Recognition
• Task-adaptive Neural Process for User Cold-Start Recommendation
• Aligning Visual Prototypes with BERT Embeddings for Few-Shot Learning
• Meta Adversarial Perturbations
• Decoder Choice Network for Meta-Learning
• Bilevel Optimization for Machine Learning: Algorithm Design and Convergence Analysis
• Intelligence, physics and information -- the tradeoff between accuracy and simplicity in machine learning
• ReMP: Rectified Metric Propagation for Few-Shot Learning
• Deep Mixed Effect Model using Gaussian Processes: A Personalized and Reliable Prediction for Healthcare
• Online Hyperparameter Meta-Learning with Hypergradient Distillation
• Few-Shot Learning with Intra-Class Knowledge Transfer
• How Does the Task Landscape Affect MAML Performance?
• Contextualizing Enhances Gradient Based Meta Learning
• EVO-RL: Evolutionary-Driven Reinforcement Learning
• Few-shot Learning for Spatial Regression
• Model-Agnostic Learning to Meta-Learn
• Stress Testing of Meta-learning Approaches for Few-shot Learning
• How to trust unlabeled data? Instance Credibility Inference for Few-Shot Learning
• Learn Faster and Forget Slower via Fast and Stable Task Adaptation
• Meta-learning for mixed linear regression
• Learning Meta-class Memory for Few-Shot Semantic Segmentation
• ST-MAML: A Stochastic-Task based Method for Task-Heterogeneous Meta-Learning
• An Optimization-Based Meta-Learning Model for MRI Reconstruction with Diverse Dataset
• Population-Based Evolution Optimizes a Meta-Learning Objective
• A Hybrid Approach with Optimization and Metric-based Meta-Learner for Few-Shot Learning
• Hierarchical Meta Learning
• Meta-Learning with Network Pruning
• Model-Agnostic Meta-Learning using Runge-Kutta Methods
• VIABLE: Fast Adaptation via Backpropagating Learned Loss
• Testing the Genomic Bottleneck Hypothesis in Hebbian Meta-Learning
• Fast Model Editing at Scale
• Federated Few-Shot Learning with Adversarial Learning
• Meta-learning One-class Classifiers with Eigenvalue Solvers for Supervised Anomaly Detection
• Multi-level Similarity Learning for Low-Shot Recognition
• Representation based and Attention augmented Meta learning
• Towards Understanding Residual and Dilated Dense Neural Networks via Convolutional Sparse Coding
• MetaAugment: Sample-Aware Data Augmentation Policy Learning
• Identifying Physical Law of Hamiltonian Systems via Meta-Learning
• Data Summarization via Bilevel Optimization
• Variable-Shot Adaptation for Online Meta-Learning
• Personalized Adaptive Meta Learning for Cold-start User Preference Prediction
• When the Few Outweigh the Many: Illicit Content Recognition with Few-Shot Learning
• GCCN: Global Context Convolutional Network
• Rapid Model Architecture Adaption for Meta-Learning
• A Few-Shot Sequential Approach for Object Counting
• Meta-Teacher For Face Anti-Spoofing
• Meta-Learning with Adjoint Methods
• FISAR: Forward Invariant Safe Reinforcement Learning with a Deep Neural Network-Based Optimize
• Relation-aware Meta-learning for Market Segment Demand Prediction with Limited Records
• Cooperative Bi-path Metric for Few-shot Learning
• Class Interference Regularization
• BOML: A Modularized Bilevel Optimization Library in Python for Meta Learning
• Is Fast Adaptation All You Need?
• Meta-Active Learning for Node Response Prediction in Graphs
• One-Class Meta-Learning: Towards Generalizable Few-Shot Open-Set Classification
• Learning by Examples Based on Multi-level Optimization
• Learning to Generalize to Unseen Tasks with Bilevel Optimization
• MetaMix: Improved Meta-Learning with Interpolation-based Consistency Regularization
• Task Attended Meta-Learning for Few-Shot Learning
• MTL2L: A Context Aware Neural Optimiser
• Meta-learning for Matrix Factorization without Shared Rows or Columns
• On Data Efficiency of Meta-learning
• Few-shot Image Classification with Multi-Facet Prototypes
• HMRL: Hyper-Meta Learning for Sparse Reward Reinforcement Learning Problem
• When Autonomous Systems Meet Accuracy and Transferability through AI: A Survey
• Learning to Continually Learn Rapidly from Few and Noisy Data
• Adaptive Policy Transfer in Reinforcement Learning
• Beyond Categorical Label Representations for Image Classification
• Towards Enabling Meta-Learning from Target Models
• Few-shot Learning for Topic Modeling
• One-Shot Image Classification by Learning to Restore Prototypes
• Meta Cyclical Annealing Schedule: A Simple Approach to Avoiding Meta-Amortization Error
• Meta-Learning with Versatile Loss Geometries for Fast Adaptation Using Mirror Descent