A Comprehensive Survey of Continual Learning: Theory, Method and Application

2023/01/31 by Liyuan Wang, Xingxing Zhang, Wang, Liyuan +5 · 126 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Multimodal Machine Learning Applications

paper · pdf · doi:10.48550/arxiv.2302.00487

openalex publication_date 2023/01/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

To cope with real-world dynamics, an intelligent system needs to incrementally acquire, update, accumulate, and exploit knowledge throughout its lifetime. This ability, known as continual learning, provides a foundation for AI systems to develop themselves adaptively. In a general sense, continual learning is explicitly limited by catastrophic forgetting, where learning a new task usually results in a dramatic performance degradation of the old tasks. Beyond this, increasingly numerous advances have emerged in recent years that largely extend the understanding and application of continual learning. The growing and widespread interest in this direction demonstrates its realistic significance as well as complexity. In this work, we present a comprehensive survey of continual learning, seeking to bridge the basic settings, theoretical foundations, representative methods, and practical applications. Based on existing theoretical and empirical results, we summarize the general objectives of continual learning as ensuring a proper stability-plasticity trade-off and an adequate intra/inter-task generalizability in the context of resource efficiency. Then we provide a state-of-the-art and elaborated taxonomy, extensively analyzing how representative methods address continual learning, and how they are adapted to particular challenges in realistic applications. Through an in-depth discussion of promising directions, we believe that such a holistic perspective can greatly facilitate subsequent exploration in this field and beyond.

Cited by

• LibContinual: A Comprehensive Library towards Realistic Continual Learning
• Rethinking Knowledge Distillation in Collaborative Machine Learning: Memory, Knowledge, and Their Interactions
• InvCoSS: Inversion-driven Continual Self-supervised Learning in Medical Multi-modal Image Pre-training
• Steering Vision-Language Pre-trained Models for Incremental Face Presentation Attack Detection
• AL-GNN: Privacy-Preserving and Replay-Free Continual Graph Learning via Analytic Learning
• Sophia: A Persistent Agent Framework of Artificial Life
• M2RU: Memristive Minion Recurrent Unit for On-Chip Continual Learning at the Edge
• Sequencing to Mitigate Catastrophic Forgetting in Continual Learning
• Out-of-Distribution Detection for Continual Learning: Design Principles and Benchmarking
• On the Dangers of Bootstrapping Generation for Continual Learning and Beyond
• Multi-Intent Spoken Language Understanding: Methods, Trends, and Challenges
• Representation Calibration and Uncertainty Guidance for Class-Incremental Learning based on Vision Language Model
• Efficient Continual Learning in Neural Machine Translation: A Low-Rank Adaptation Approach
• Robust Finetuning of Vision-Language-Action Robot Policies via Parameter Merging
• Multi-Generator Continual Learning for Robust Delay Prediction in 6G
• Weighted Contrastive Learning for Anomaly-Aware Time-Series Forecasting
• Vision and Causal Learning Based Channel Estimation for THz Communications
• Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models
• Provably Safe Model Updates
• Memory-Integrated Reconfigurable Adapters: A Unified Framework for Settings with Multiple Tasks
• Memory-Amortized Inference: A Topological Unification of Search, Closure, and Structure
• The Geometry of Certainty: Recursive Topological Condensation and the Limits of Inference
• Embodied Intelligent Wireless (EIW): Synesthesia of Machines Empowered Wireless Communications
• Knowledge Distillation for Continual Learning of Biomedical Neural Fields
• Harmonious Parameter Adaptation in Continual Visual Instruction Tuning for Safety-Aligned MLLMs
• ModHiFi: Identifying High Fidelity predictive components for Model Modification
• Adversarial Pseudo-replay for Exemplar-free Class-incremental Learning
• Intrinsic preservation of plasticity in continual quantum learning
• Hierarchical Semantic Tree Anchoring for CLIP-Based Class-Incremental Learning
• Learning from Mistakes: Loss-Aware Memory Enhanced Continual Learning for LiDAR Place Recognition
• Multimodal Continual Instruction Tuning with Dynamic Gradient Guidance
• AnaCP: Toward Upper-Bound Continual Learning via Analytic Contrastive Projection
• Dual-LoRA and Quality-Enhanced Pseudo Replay for Multimodal Continual Food Learning
• Online Continual Learning on Intel Loihi 2 via a Co-designed Spiking Neural Network
• Retrofit: Continual Learning with Bounded Forgetting for Security Applications
• Data Heterogeneity and Forgotten Labels in Split Federated Learning
• Efficient Model-Agnostic Continual Learning for Next POI Recommendation
• Continual Unlearning for Text-to-Image Diffusion Models: A Regularization Perspective
• Operational machine learning for remote spectroscopic detection of CH₄ point sources
• Let's Split Up: Zero-Shot Classifier Edits for Fine-Grained Video Understanding
• Forgetting is Everywhere
• The brain as a blueprint: a survey of brain-inspired approaches to learning in artificial intelligence
• A Feedback-Control Framework for Efficient Dataset Collection from In-Vehicle Data Streams
• In Situ Training of Implicit Neural Compressors for Scientific Simulations via Sketch-Based Regularization
• Efficient Online Continual Learning in Sensor-Based Human Activity Recognition
• What's the next frontier for Data-centric AI? Data Savvy Agents
• Parameterized Prompt for Incremental Object Detection
• Can Vision-Language-Action Models Learn from Real-World Data Continually without Forgetting?
• Gated Adaptation for Continual Learning in Human Activity Recognition
• MemEIC: A Step Toward Continual and Compositional Knowledge Editing
• Explaining Robustness to Catastrophic Forgetting Through Incremental Concept Formation
• OFFSIDE: Benchmarking Unlearning Misinformation in Multimodal Large Language Models
• Time-varying Gaussian Process Bandit Optimization with Experts: no-regret in logarithmically-many side queries
• Model Merging with Functional Dual Anchors
• PLAN: Proactive Low-Rank Allocation for Continual Learning
• More Than Memory Savings: Zeroth-Order Optimization Mitigates Forgetting in Continual Learning
• CO-PFL: Contribution-Oriented Personalized Federated Learning for Heterogeneous Networks
• Information Theory in Open-world Machine Learning Foundations, Frameworks, and Future Direction
• Mapping Post-Training Forgetting in Language Models at Scale
• Beyond Binary Out-of-Distribution Detection: Characterizing Distributional Shifts with Multi-Statistic Diffusion Trajectories
• CaMiT: A Time-Aware Car Model Dataset for Classification and Generation
• Domain Generalizable Continual Learning
• STABLE: Gated Continual Learning for Large Language Models
• Time-Varying Optimization for Streaming Data Via Temporal Weighting
• End-to-End Test-Time Training for Long Context
• Continual Learning for Image Captioning through Improved Image-Text Alignment
• IMLP: An Energy-Efficient Continual Learning Method for Tabular Data Streams
• AgentTypo: Adaptive Typographic Prompt Injection Attacks against Black-box Multimodal Agents
• Using predefined vector systems as latent space configuration for neural network supervised training on data with arbitrarily large number of classes
• Deep Generative Continual Learning using Functional LoRA: FunLoRA
• Rehearsal-free and Task-free Online Continual Learning With Contrastive Prompt
• Continual Learning with Query-Only Attention
• Foundation vs. Specialized Models: Evaluating Catastrophic Forgetting in Continual Time Series Forecasting
• Performance-Efficiency Trade-off for Fashion Image Retrieval
• FS-KAN: Permutation Equivariant Kolmogorov-Arnold Networks via Function Sharing
• Continual Learning to Generalize Forwarding Strategies for Diverse Mobile Wireless Networks
• Toward a Holistic Approach to Continual Model Merging
• CLAD-Net: Continual Activity Recognition in Multi-Sensor Wearable Systems
• Hierarchical Representation Matching for CLIP-based Class-Incremental Learning
• The Lie of the Average: How Class Incremental Learning Evaluation Deceives You?
• Lifelong Learning with Behavior Consolidation for Vehicle Routing
• LANCE: Low Rank Activation Compression for Efficient On-Device Continual Learning
• SFT Doesn't Always Hurt General Capabilities: Revisiting Domain-Specific Fine-Tuning in LLMs
• Sig2Model: A Boosting-Driven Model for Updatable Learned Indexes
• Adaptive Model Ensemble for Continual Learning
• End-to-end deep attention-based multitask pipeline for predicting uncertainty-quantified peptide properties from mass spectrometry data
• Choice Outweighs Effort: Facilitating Complementary Knowledge Fusion in Federated Learning via Re-calibration and Merit-discrimination
• COLT: Enhancing Video Large Language Models with Continual Tool Usage
• RoboSeek: You Need to Interact with Your Objects
• CBPNet: A Continual Backpropagation Prompt Network for Alleviating Plasticity Loss on Edge Devices
• AFT: An Exemplar-Free Class Incremental Learning Method for Environmental Sound Classification
• EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer
• FedTeddi: Temporal Drift and Divergence Aware Scheduling for Timely Federated Edge Learning
• Genesis: A Spiking Neuromorphic Accelerator With On-chip Continual Learning
• Artificial intelligence for representing and characterizing quantum systems
• Seeing Clearly, Forgetting Deeply: Revisiting Fine-Tuned Video Generators for Driving Simulation
• RL's Razor: Why Online Reinforcement Learning Forgets Less
• Online time series prediction using feature adjustment
• BM-CL: Bias Mitigation through the lens of Continual Learning
• Adapting to Change: A Comparison of Continual and Transfer Learning for Modeling Building Thermal Dynamics under Concept Drifts
• Complementary Learning System Empowers Online Continual Learning of Vehicle Motion Forecasting in Smart Cities
• CITADEL: Continual Anomaly Detection for Enhanced Learning in IoT Intrusion Detection
• High-dimensional Asymptotics of Generalization Performance in Continual Ridge Regression
• C-Flat++: Towards a More Efficient and Powerful Framework for Continual Learning
• Incremental Object Detection with Prompt-based Methods
• RICO: Two Realistic Benchmarks and an In-Depth Analysis for Incremental Learning in Object Detection
• Monte Carlo Functional Regularisation for Continual Learning
• Omni Survey for Multimodality Analysis in Visual Object Tracking
• Data Shift of Object Detection in Autonomous Driving
• Towards Efficient Prompt-based Continual Learning in Distributed Medical AI
• Enhancing Memory Recall in LLMs with Gauss-Tin: A Hybrid Instructional and Gaussian Replay Approach
• Exploring Cross-Stage Adversarial Transferability in Class-Incremental Continual Learning
• Multi-level Collaborative Distillation Meets Global Workspace Model: A Unified Framework for OCIL
• On Understanding of the Dynamics of Model Capacity in Continual Learning
• OpenHAIV: A Framework Towards Practical Open-World Learning
• MCITlib: Multimodal Continual Instruction Tuning Library and Benchmark
• Lifelong Learner: Discovering Versatile Neural Solvers for Vehicle Routing Problems
• Divide-and-Conquer for Enhancing Unlabeled Learning, Stability, and Plasticity in Semi-supervised Continual Learning
• GeRe: Towards Efficient Anti-Forgetting in Continual Learning of LLM via General Samples Replay
• Revisiting Continual Semantic Segmentation with Pre-trained Vision Models
• GaitAdapt: Continual Learning for Evolving Gait Recognition
• Exploring Stability-Plasticity Trade-offs for Continual Named Entity Recognition
• FedPromo: Federated Lightweight Proxy Models at the Edge Bring New Domains to Foundation Models
• Data-driven RF Tomography via Cross-modal Sensing and Continual Learning
• Continual Learning with Synthetic Boundary Experience Blending
• Forgetting of task-specific knowledge in model merging-based continual learning

Related

• Continual Learning as Computationally Constrained Reinforcement Learning
• Reviewing continual learning from the perspective of human-level intelligence
• Rethinking Continual Learning for Autonomous Agents and Robots
• Continual Learning and Catastrophic Forgetting
• Continual Learning in Neural Networks