A continual learning survey: Defying forgetting in classification tasks

Abstract

Artificial neural networks thrive in solving the classification problem for a particular rigid task, acquiring knowledge through generalized learning behaviour from a distinct training phase. The resulting network resembles a static entity of knowledge, with endeavours to extend this knowledge without targeting the original task resulting in a catastrophic forgetting. Continual learning shifts this paradigm towards networks that can continually accumulate knowledge over different tasks without the need to retrain from scratch. We focus on task incremental classification, where tasks arrive sequentially and are delineated by clear boundaries. Our main contributions concern 1) a taxonomy and extensive overview of the state-of-the-art, 2) a novel framework to continually determine the stability-plasticity trade-off of the continual learner, 3) a comprehensive experimental comparison of 11 state-of-the-art continual learning methods and 4 baselines. We empirically scrutinize method strengths and weaknesses on three benchmarks, considering Tiny Imagenet and large-scale unbalanced iNaturalist and a sequence of recognition datasets. We study the influence of model capacity, weight decay and dropout regularization, and the order in which the tasks are presented, and qualitatively compare methods in terms of required memory, computation time, and storage.

Accepted TPAMI paper, including Appendix, code publicly available

References in corpus (42)

• Distilling the Knowledge in a Neural Network
• Overcoming catastrophic forgetting in neural networks
• Improving neural networks by preventing co-adaptation of feature detectors
• Fine-Grained Visual Classification of Aircraft
• Progressive Neural Networks
• Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence
• Continual Learning Through Synaptic Intelligence
• An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networks
• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• A Survey on Multi-Task Learning
• Efficient Lifelong Learning with A-GEM
• Towards Open World Recognition
• Experience Replay for Continual Learning
• On Tiny Episodic Memories in Continual Learning
• New insights and perspectives on the natural gradient method
• Progress & Compress: A scalable framework for continual learning
• Variational Continual Learning
• Overcoming Catastrophic Forgetting by Incremental Moment Matching
• Overcoming catastrophic forgetting with hard attention to the task
• Encoder Based Lifelong Learning
• Re-evaluating Continual Learning Scenarios: A Categorization and Case for Strong Baselines
• Qualitatively characterizing neural network optimization problems
• Reinforced Continual Learning
• Measuring Catastrophic Forgetting in Neural Networks
• Towards Robust Evaluations of Continual Learning
• Less-forgetting Learning in Deep Neural Networks
• Revisiting Natural Gradient for Deep Networks
• A PAC-Bayesian bound for Lifelong Learning
• Uncertainty-based Continual Learning with Adaptive Regularization
• Lifelong Generative Modeling
• On Quadratic Penalties in Elastic Weight Consolidation
• A comprehensive, application-oriented study of catastrophic forgetting in DNNs
• Open-world Learning and Application to Product Classification
• Gradient based sample selection for online continual learning
• Selfless Sequential Learning
• Task Agnostic Continual Learning Using Online Variational Bayes
• Pseudo-Recursal: Solving the Catastrophic Forgetting Problem in Deep Neural Networks
• A Unifying Bayesian View of Continual Learning
• Toward Understanding Catastrophic Forgetting in Continual Learning
• Improving and Understanding Variational Continual Learning
• Neural Topic Modeling with Continual Lifelong Learning
• Continual Classification Learning Using Generative Models

Cited by in corpus (170)

• Towards Personalized Federated Learning
• Class-Incremental Learning: A Survey
• Adaptive Aggregation Networks for Class-Incremental Learning
• The Multi-modality Cell Segmentation Challenge: Towards Universal Solutions
• Class-Incremental Continual Learning into the eXtended DER-verse
• Continual Learning for Recurrent Neural Networks: an Empirical Evaluation
• Federated Continual Learning via Knowledge Fusion: A Survey
• Temporal Knowledge Graph Completion: A Survey
• Rethinking Search: Making Domain Experts out of Dilettantes
• Adaptation Strategies for Automated Machine Learning on Evolving Data
• Dark Experience for General Continual Learning: a Strong, Simple Baseline
• A TinyML Platform for On-Device Continual Learning with Quantized Latent Replays
• Incorporating Neuro-Inspired Adaptability for Continual Learning in Artificial Intelligence
• General Purpose Artificial Intelligence Systems (GPAIS): Properties, Definition, Taxonomy, Societal Implications and Responsible Governance
• Continual Learning for Predictive Maintenance: Overview and Challenges
• Continual Learning in Sensor-based Human Activity Recognition: an Empirical Benchmark Analysis
• Towards Federated Learning against Noisy Labels via Local Self-Regularization
• DualNet: Continual Learning, Fast and Slow
• The Internet of Federated Things (IoFT): A Vision for the Future and In-depth Survey of Data-driven Approaches for Federated Learning
• ReduNet: A White-box Deep Network from the Principle of Maximizing Rate Reduction
• A Survey on All-in-One Image Restoration: Taxonomy, Evaluation and Future Trends
• Machine Learning Techniques for Sensor-based Human Activity Recognition with Data Heterogeneity -- A Review
• Continual Learning and Catastrophic Forgetting
• Subspace Regularizers for Few-Shot Class Incremental Learning
• A Survey on Continual Semantic Segmentation: Theory, Challenge, Method and Application
• Learning without Forgetting for Vision-Language Models
• Orthogonal Gradient Descent for Continual Learning
• Continual Learning for Generative Retrieval over Dynamic Corpora
• Continual Semi-Supervised Learning through Contrastive Interpolation Consistency
• AI in Human-computer Gaming: Techniques, Challenges and Opportunities
• Lifelong Continual Learning for Anomaly Detection: New Challenges, Perspectives, and Insights
• Multi-Task Incremental Learning for Object Detection
• A distillation-based approach integrating continual learning and federated learning for pervasive services
• CPR: Classifier-Projection Regularization for Continual Learning
• An Investigation of Replay-based Approaches for Continual Learning
• Federated Continual Learning: Concepts, Challenges, and Solutions
• Inherit with Distillation and Evolve with Contrast: Exploring Class Incremental Semantic Segmentation Without Exemplar Memory
• Meta-Consolidation for Continual Learning
• Tackling Noisy Clients in Federated Learning with End-to-end Label Correction
• RATT: Recurrent Attention to Transient Tasks for Continual Image Captioning
• Evolutionary Multitask Optimization: Fundamental Research Questions, Practices, and Directions for the Future
• Self-Supervised Training Enhances Online Continual Learning
• AdaptCL: Adaptive Continual Learning for Tackling Heterogeneity in Sequential Datasets
• Lifelong Adaptive Machine Learning for Sensor-based Human Activity Recognition Using Prototypical Networks
• Learning to Continuously Optimize Wireless Resource In Episodically Dynamic Environment
• CL-CaGAN: Capsule differential adversarial continuous learning for cross-domain hyperspectral anomaly detection
• Replay in Deep Learning: Current Approaches and Missing Biological Elements
• Continual learning of longitudinal health records
• Using Adapters to Overcome Catastrophic Forgetting in End-to-End Automatic Speech Recognition
• Learnware: Small Models Do Big
• Continual World: A Robotic Benchmark For Continual Reinforcement Learning
• When Meta-Learning Meets Online and Continual Learning: A Survey
• AirLoop: Lifelong Loop Closure Detection
• Continuous fake media detection: adapting deepfake detectors to new generative techniques
• Multi-View Class Incremental Learning
• Complementary Calibration: Boosting General Continual Learning with Collaborative Distillation and Self-Supervision
• Catastrophic Interference in Reinforcement Learning: A Solution Based on Context Division and Knowledge Distillation
• Continual Prune-and-Select: Class-incremental learning with specialized subnetworks
• A Domain-Agnostic Approach for Characterization of Lifelong Learning Systems
• New Insights on Relieving Task-Recency Bias for Online Class Incremental Learning
• Evaluating Logical Generalization in Graph Neural Networks
• Continual Horizontal Federated Learning for Heterogeneous Data
• Online Continual Learning via the Knowledge Invariant and Spread-out Properties
• Class Gradient Projection For Continual Learning
• AbdomenCT-1K: Is Abdominal Organ Segmentation A Solved Problem?
• A Neural Span-Based Continual Named Entity Recognition Model
• Multicenter Privacy-Preserving Model Training for Deep Learning Brain Metastases Autosegmentation
• Personalized Federated Continual Learning via Multi-granularity Prompt
• Exploring the Capabilities and Limitations of Large Language Models for Radiation Oncology Decision Support
• A multifidelity approach to continual learning for physical systems
• Advances in Continual Graph Learning for Anti-Money Laundering Systems: A Comprehensive Review
• Dynamic Sub-graph Distillation for Robust Semi-supervised Continual Learning
• Task Relation-aware Continual User Representation Learning
• Processing Natural Language on Embedded Devices: How Well Do Transformer Models Perform?
• DSLR: Diversity Enhancement and Structure Learning for Rehearsal-based Graph Continual Learning
• Supervised Contrastive Replay: Revisiting the Nearest Class Mean Classifier in Online Class-Incremental Continual Learning
• Generative Feature Replay For Class-Incremental Learning
• Encoders and Ensembles for Task-Free Continual Learning
• Continual Learning for Monolingual End-to-End Automatic Speech Recognition
• QCore: Data-Efficient, On-Device Continual Calibration for Quantized Models -- Extended Version
• On Class Orderings for Incremental Learning
• cPNN: Continuous Progressive Neural Networks for Evolving Streaming Time Series
• Half-Real Half-Fake Distillation for Class-Incremental Semantic Segmentation
• A Survey on Curriculum Learning
• Tackling Dynamics in Federated Incremental Learning with Variational Embedding Rehearsal
• Continual Learning with Neuromorphic Computing: Foundations, Methods, and Emerging Applications
• Exploring the Intersection between Neural Architecture Search and Continual Learning
• Latent Spectral Regularization for Continual Learning
• Weight Averaging: A Simple Yet Effective Method to Overcome Catastrophic Forgetting in Automatic Speech Recognition
• A Self-Commissioning Edge Computing Method for Data-Driven Anomaly Detection in Power Electronic Systems
• On Sequential Bayesian Inference for Continual Learning
• Broadening AI Ethics Narratives: An Indic Art View
• CL-BioGAN: Biologically-Inspired Cross-Domain Continual Learning for Hyperspectral Anomaly Detection
• Continual Model-Based Reinforcement Learning with Hypernetworks
• On the importance of cross-task features for class-incremental learning
• Training for the Future: A Simple Gradient Interpolation Loss to Generalize Along Time
• A Study of Continual Learning Methods for Q-Learning
• Contrastive Supervised Distillation for Continual Representation Learning
• Hessian Aware Low-Rank Perturbation for Order-Robust Continual Learning
• Continual Learning Using Bayesian Neural Networks
• Towards Fast and Stable Federated Learning: Confronting Heterogeneity via Knowledge Anchor
• Re-using Adversarial Mask Discriminators for Test-time Training under Distribution Shifts
• Positive-Congruent Training: Towards Regression-Free Model Updates
• CL2R: Compatible Lifelong Learning Representations
• Towards Better Plasticity-Stability Trade-off in Incremental Learning: A Simple Linear Connector
• Continual Learning with Fully Probabilistic Models
• Online Continual Learning with Natural Distribution Shifts: An Empirical Study with Visual Data
• Confidence-Guided Learning Process for Continuous Classification of Time Series
• Look At Me, No Replay! SurpriseNet: Anomaly Detection Inspired Class Incremental Learning
• Continual Learning in Predictive Autoscaling
• VIPeR: Visual Incremental Place Recognition with Adaptive Mining and Continual Learning
• MEIL-NeRF: Memory-Efficient Incremental Learning of Neural Radiance Fields
• Sequoia: A Software Framework to Unify Continual Learning Research
• Continual Learning in the Presence of Repetition
• Lifelong Vehicle Trajectory Prediction Framework Based on Generative Replay
• Memory-Free Generative Replay For Class-Incremental Learning
• Co-Transport for Class-Incremental Learning
• SLAW: Scaled Loss Approximate Weighting for Efficient Multi-Task Learning
• AdaHAT: Adaptive Hard Attention to the Task in Task-Incremental Learning
• I2CANSAY:Inter-Class Analogical Augmentation and Intra-Class Significance Analysis for Non-Exemplar Online Task-Free Continual Learning
• Reducing Geographic Disparities in Automatic Speech Recognition via Elastic Weight Consolidation
• One-Shot Domain Incremental Learning
• Optimal Protocols for Continual Learning via Statistical Physics and Control Theory
• CLEVA-Compass: A Continual Learning EValuation Assessment Compass to Promote Research Transparency and Comparability
• Disentanglement of Color and Shape Representations for Continual Learning
• Approximating Opaque Top-k Queries
• FETCH: A Memory-Efficient Replay Approach for Continual Learning in Image Classification
• System Design for an Integrated Lifelong Reinforcement Learning Agent for Real-Time Strategy Games
• DAIR: Disentangled Attention Intrinsic Regularization for Safe and Efficient Bimanual Manipulation
• AlterSGD: Finding Flat Minima for Continual Learning by Alternative Training
• Curriculum learning for multilevel budgeted combinatorial problems
• Continual Learning in Task-Oriented Dialogue Systems
• A Conceptual Framework for Lifelong Learning
• Guiding the retraining of convolutional neural networks against adversarial inputs
• Selective Replay Enhances Learning in Online Continual Analogical Reasoning
• When Video Classification Meets Incremental Classes
• A Frank System for Co-Evolutionary Hybrid Decision-Making
• Tuned Compositional Feature Replays for Efficient Stream Learning
• Unsupervised Class-Incremental Learning Through Confusion
• DeCoR: Defy Knowledge Forgetting by Predicting Earlier Audio Codes
• Unsupervised Continual Learning Via Pseudo Labels
• Continual Learning With Quasi-Newton Methods
• A Procedural World Generation Framework for Systematic Evaluation of Continual Learning
• Juvenile state hypothesis: What we can learn from lottery ticket hypothesis researches?
• Expanding continual few-shot learning benchmarks to include recognition of specific instances
• MyMigrationBot: A Cloud-based Facebook Social Chatbot for Migrant Populations
• Accelerating Distributed Online Meta-Learning via Multi-Agent Collaboration under Limited Communication
• Statistical Mechanical Analysis of Catastrophic Forgetting in Continual Learning with Teacher and Student Networks
• RSAC: Regularized Subspace Approximation Classifier for Lightweight Continuous Learning
• Continual Learning of Generative Models with Limited Data: From Wasserstein-1 Barycenter to Adaptive Coalescence
• Learning to Transfer: A Foliated Theory
• Continually Learn to Map Visual Concepts to Large Language Models in Resource-constrained Environments
• Context selectivity with dynamic availability enables lifelong continual learning
• What Affects the Performance of Fake Audio Detection? Analyzing Factors in a Continual Learning Setting
• Understanding the Dynamics of DNNs Using Graph Modularity
• Replay4NCL: An Efficient Memory Replay-based Methodology for Neuromorphic Continual Learning in Embedded AI Systems
• Benchmarking Catastrophic Forgetting Mitigation Methods in Federated Time Series Forecasting
• Inverse-Hessian Regularization for Continual Learning in ASR
• Online Domain-aware LLM Decoding for Continual Domain Evolution
• DeCoFlow: Structural Decomposition of Normalizing Flows for Continual Anomaly Detection
• Unleashing the Potential of Large Language Models: A Blueprint for Real-Time, Enterprise-Ready Deployments
• PARMESAN: Parameter-Free Memory Search and Transduction for Dense Prediction Tasks
• Don't Look Back in Anger: MAGIC Net for Streaming Continual Learning with Temporal Dependence
• Adaptive Explainable Continual Learning Framework for Regression Problems with Focus on Power Forecasts
• MAcPNN: Mutual Assisted Learning on Data Streams with Temporal Dependence
• Divide and Conquer: Static-Dynamic Collaboration for Few-Shot Class-Incremental Learning
• Continual Novelty Detection
• Regression Bugs Are In Your Model! Measuring, Reducing and Analyzing Regressions In NLP Model Updates
• TAG: Task-based Accumulated Gradients for Lifelong learning
• Continual Self-Supervised Learning with Masked Autoencoders in Remote Sensing