Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence

Abstract

Incremental learning (IL) has received a lot of attention recently, however, the literature lacks a precise problem definition, proper evaluation settings, and metrics tailored specifically for the IL problem. One of the main objectives of this work is to fill these gaps so as to provide a common ground for better understanding of IL. The main challenge for an IL algorithm is to update the classifier whilst preserving existing knowledge. We observe that, in addition to forgetting, a known issue while preserving knowledge, IL also suffers from a problem we call intransigence, inability of a model to update its knowledge. We introduce two metrics to quantify forgetting and intransigence that allow us to understand, analyse, and gain better insights into the behaviour of IL algorithms. We present RWalk, a generalization of EWC++ (our efficient version of EWC [Kirkpatrick2016EWC]) and Path Integral [Zenke2017Continual] with a theoretically grounded KL-divergence based perspective. We provide a thorough analysis of various IL algorithms on MNIST and CIFAR-100 datasets. In these experiments, RWalk obtains superior results in terms of accuracy, and also provides a better trade-off between forgetting and intransigence.

References in corpus (16)

• Adam: A Method for Stochastic Optimization
• Distilling the Knowledge in a Neural Network
• Overcoming catastrophic forgetting in neural networks
• Deep Residual Learning for Image Recognition
• Progressive Neural Networks
• Continual Learning Through Synaptic Intelligence
• Learning multiple visual domains with residual adapters
• Gradient Episodic Memory for Continual Learning
• Optimizing Neural Networks with Kronecker-factored Approximate Curvature
• Progress & Compress: A scalable framework for continual learning
• Variational Continual Learning
• Overcoming Catastrophic Forgetting by Incremental Moment Matching
• Continual Learning with Deep Generative Replay
• Lifelong Learning with Dynamically Expandable Networks
• Revisiting Natural Gradient for Deep Networks
• A Kronecker-factored approximate Fisher matrix for convolution layers

Cited by in corpus (185)

• A continual learning survey: Defying forgetting in classification tasks
• Three scenarios for continual learning
• On Tiny Episodic Memories in Continual Learning
• Class-Incremental Learning: A Survey
• Adaptive Aggregation Networks for Class-Incremental Learning
• Towards Robust Evaluations of Continual Learning
• Compacting, Picking and Growing for Unforgetting Continual Learning
• Class-Incremental Continual Learning into the eXtended DER-verse
• Continual Learning for Recurrent Neural Networks: an Empirical Evaluation
• Learning to Learn without Forgetting by Maximizing Transfer and Minimizing Interference
• Knowledge-Preserving Incremental Social Event Detection via Heterogeneous GNNs
• Understanding the Role of Training Regimes in Continual Learning
• Dark Experience for General Continual Learning: a Strong, Simple Baseline
• Incorporating Neuro-Inspired Adaptability for Continual Learning in Artificial Intelligence
• Uncertainty-aware Contrastive Distillation for Incremental Semantic Segmentation
• Online Continual Learning with Maximally Interfered Retrieval
• Selfless Sequential Learning
• Task Agnostic Continual Learning Using Online Variational Bayes
• Memorizing Complementation Network for Few-Shot Class-Incremental Learning
• Continual Learning in Sensor-based Human Activity Recognition: an Empirical Benchmark Analysis
• DualNet: Continual Learning, Fast and Slow
• Continual Deep Learning by Functional Regularisation of Memorable Past
• Overcoming Long-term Catastrophic Forgetting through Adversarial Neural Pruning and Synaptic Consolidation
• Triple Memory Networks: a Brain-Inspired Method for Continual Learning
• A Unifying Bayesian View of Continual Learning
• Cognitively-Inspired Model for Incremental Learning Using a Few Examples
• Continual Learning on Dynamic Graphs via Parameter Isolation
• A Survey on Continual Semantic Segmentation: Theory, Challenge, Method and Application
• M2KD: Multi-model and Multi-level Knowledge Distillation for Incremental Learning
• DER: Dynamically Expandable Representation for Class Incremental Learning
• CLeaR: An Adaptive Continual Learning Framework for Regression Tasks
• Recent Advances of Continual Learning in Computer Vision: An Overview
• Learning without Forgetting for Vision-Language Models
• Unified Probabilistic Deep Continual Learning through Generative Replay and Open Set Recognition
• Towards Continual Egocentric Activity Recognition: A Multi-modal Egocentric Activity Dataset for Continual Learning
• Rotate your Networks: Better Weight Consolidation and Less Catastrophic Forgetting
• Multi-Task Incremental Learning for Object Detection
• An Investigation of Replay-based Approaches for Continual Learning
• CPR: Classifier-Projection Regularization for Continual Learning
• Toward Understanding Catastrophic Forgetting in Continual Learning
• Federated Continual Learning: Concepts, Challenges, and Solutions
• Inherit with Distillation and Evolve with Contrast: Exploring Class Incremental Semantic Segmentation Without Exemplar Memory
• Continual Learning in Low-rank Orthogonal Subspaces
• Improving and Understanding Variational Continual Learning
• Hierarchical Memory Learning for Fine-Grained Scene Graph Generation
• Class-incremental Learning via Deep Model Consolidation
• Improved Schemes for Episodic Memory-based Lifelong Learning
• Continual Learning with Adaptive Weights (CLAW)
• RODEO: Replay for Online Object Detection
• Meta-Consolidation for Continual Learning
• Continual Learning with Node-Importance based Adaptive Group Sparse Regularization
• RATT: Recurrent Attention to Transient Tasks for Continual Image Captioning
• EEC: Learning to Encode and Regenerate Images for Continual Learning
• Lifelong Object Detection
• Lifelong Adaptive Machine Learning for Sensor-based Human Activity Recognition Using Prototypical Networks
• Self-Supervised Training Enhances Online Continual Learning
• Boosting Deep Open World Recognition by Clustering
• Replay in Deep Learning: Current Approaches and Missing Biological Elements
• CL-CaGAN: Capsule differential adversarial continuous learning for cross-domain hyperspectral anomaly detection
• REMIND Your Neural Network to Prevent Catastrophic Forgetting
• Cost-effective On-device Continual Learning over Memory Hierarchy with Miro
• When Meta-Learning Meets Online and Continual Learning: A Survey
• AirLoop: Lifelong Loop Closure Detection
• Online Lifelong Generalized Zero-Shot Learning
• Rainbow Memory: Continual Learning with a Memory of Diverse Samples
• Distilling Causal Effect of Data in Class-Incremental Learning
• New Insights on Relieving Task-Recency Bias for Online Class Incremental Learning
• An Adaptive Random Path Selection Approach for Incremental Learning
• Mesoscopic description of hippocampal replay and metastability in spiking neural networks with short-term plasticity
• Continual Prune-and-Select: Class-incremental learning with specialized subnetworks
• Task Agnostic Continual Learning Using Online Variational Bayes with Fixed-Point Updates
• Topology-aware Embedding Memory for Continual Learning on Expanding Networks
• Learning at a Glance: Towards Interpretable Data-limited Continual Semantic Segmentation via Semantic-Invariance Modelling
• Meta-Learning with Sparse Experience Replay for Lifelong Language Learning
• Carousel Memory: Rethinking the Design of Episodic Memory for Continual Learning
• Semantic Drift Compensation for Class-Incremental Learning
• Modeling the Background for Incremental and Weakly-Supervised Semantic Segmentation
• Online Continual Learning via the Knowledge Invariant and Spread-out Properties
• GP-Tree: A Gaussian Process Classifier for Few-Shot Incremental Learning
• Online Class-Incremental Continual Learning with Adversarial Shapley Value
• Online Fast Adaptation and Knowledge Accumulation: a New Approach to Continual Learning
• Continual Learning in Neural Networks
• Linear Mode Connectivity in Multitask and Continual Learning
• Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis
• Tackling Catastrophic Forgetting and Background Shift in Continual Semantic Segmentation
• Encoders and Ensembles for Task-Free Continual Learning
• Supervised Contrastive Replay: Revisiting the Nearest Class Mean Classifier in Online Class-Incremental Continual Learning
• Recall and Learn: Fine-tuning Deep Pretrained Language Models with Less Forgetting
• Online Deep Learning based on Auto-Encoder
• Hierarchical Indian Buffet Neural Networks for Bayesian Continual Learning
• Generative Feature Replay For Class-Incremental Learning
• Training Networks in Null Space of Feature Covariance for Continual Learning
• Incremental Learning with Maximum Entropy Regularization: Rethinking Forgetting and Intransigence
• Generalized Continual Zero-Shot Learning
• Generative Replay-based Continual Zero-Shot Learning
• Learning Adaptive Embedding Considering Incremental Class
• On Class Orderings for Incremental Learning
• Half-Real Half-Fake Distillation for Class-Incremental Semantic Segmentation
• CORA: Benchmarks, Baselines, and Metrics as a Platform for Continual Reinforcement Learning Agents
• Gradient Episodic Memory with a Soft Constraint for Continual Learning
• Imbalanced Continual Learning with Partitioning Reservoir Sampling
• Lifelong Graph Learning
• Rethinking Knowledge Distillation in Collaborative Machine Learning: Memory, Knowledge, and Their Interactions
• On the Usage of Continual Learning for Out-of-Distribution Generalization in Pre-trained Language Models of Code
• Continual Learning with Neuromorphic Computing: Foundations, Methods, and Emerging Applications
• ARCADe: A Rapid Continual Anomaly Detector
• Online Structured Meta-learning
• Class Balance Matters to Active Class-Incremental Learning
• Prototype-based Incremental Few-Shot Semantic Segmentation
• Online Distillation with Continual Learning for Cyclic Domain Shifts
• Storing Encoded Episodes as Concepts for Continual Learning
• Continual Learning through Human-Robot Interaction: Human Perceptions of a Continual Learning Robot in Repeated Interactions
• Continuous Coordination As a Realistic Scenario for Lifelong Learning
• A Study of Continual Learning Methods for Q-Learning
• On the importance of cross-task features for class-incremental learning
• BI-MAML: Balanced Incremental Approach for Meta Learning
• Federated Learning and catastrophic forgetting in pervasive computing: demonstration in HAR domain
• Unifying Regularisation Methods for Continual Learning
• Online Continual Learning on Class Incremental Blurry Task Configuration with Anytime Inference
• Positive-Congruent Training: Towards Regression-Free Model Updates
• Multi-Domain Multi-Task Rehearsal for Lifelong Learning
• MEIL-NeRF: Memory-Efficient Incremental Learning of Neural Radiance Fields
• Layerwise Optimization by Gradient Decomposition for Continual Learning
• Towards Better Plasticity-Stability Trade-off in Incremental Learning: A Simple Linear Connector
• Continual Learning for Text Classification with Information Disentanglement Based Regularization
• Learnable Expansion-and-Compression Network for Few-shot Class-Incremental Learning
• Lifelong Learning with Sketched Structural Regularization
• Rethinking Experience Replay: a Bag of Tricks for Continual Learning
• A Deep Learning Framework for Lifelong Machine Learning
• Memory-Free Generative Replay For Class-Incremental Learning
• Frosting Weights for Better Continual Training
• GROWN: GRow Only When Necessary for Continual Learning
• PKI: Prior Knowledge-Infused Neural Network for Few-Shot Class-Incremental Learning
• Single-Net Continual Learning with Progressive Segmented Training (PST)
• Bilevel Continual Learning
• Reviewing continual learning from the perspective of human-level intelligence
• Continual Learning: Tackling Catastrophic Forgetting in Deep Neural Networks with Replay Processes
• Memory Efficient Class-Incremental Learning for Image Classification
• AlterSGD: Finding Flat Minima for Continual Learning by Alternative Training
• CLEVA-Compass: A Continual Learning EValuation Assessment Compass to Promote Research Transparency and Comparability
• Graph-Based Continual Learning
• Variational Auto-Regressive Gaussian Processes for Continual Learning
• Understanding Catastrophic Forgetting and Remembering in Continual Learning with Optimal Relevance Mapping
• Few-shot Continual Learning: a Brain-inspired Approach
• A Conceptual Framework for Lifelong Learning
• Learning Continually from Low-shot Data Stream
• Towards a practical measure of interference for reinforcement learning
• Revisiting Clustering of Neural Bandits: Selective Reinitialization for Mitigating Loss of Plasticity
• TIE: A Framework for Embedding-based Incremental Temporal Knowledge Graph Completion
• Learning Neural Models for Natural Language Processing in the Face of Distributional Shift
• Efficient and Robust Continual Graph Learning for Graph Classification in Biology
• Selective Replay Enhances Learning in Online Continual Analogical Reasoning
• How do Quadratic Regularizers Prevent Catastrophic Forgetting: The Role of Interpolation
• SERIL: Noise Adaptive Speech Enhancement using Regularization-based Incremental Learning
• Generative Feature Replay with Orthogonal Weight Modification for Continual Learning
• Improving Forward Compatibility in Class Incremental Learning by Increasing Representation Rank and Feature Richness
• Continual Learning With Quasi-Newton Methods
• A Procedural World Generation Framework for Systematic Evaluation of Continual Learning
• Bilevel Continual Learning
• When Video Classification Meets Incremental Classes
• Towards Recognizing New Semantic Concepts in New Visual Domains
• Understanding Adversarial Behavior of DNNs by Disentangling Non-Robust and Robust Components in Performance Metric
• Piggyback GAN: Efficient Lifelong Learning for Image Conditioned Generation
• Incremental Cross-Domain Adaptation for Robust Retinopathy Screening via Bayesian Deep Learning
• Bilevel Online Deep Learning in Non-stationary Environment
• MyMigrationBot: A Cloud-based Facebook Social Chatbot for Migrant Populations
• Split-and-Bridge: Adaptable Class Incremental Learning within a Single Neural Network
• A study on the plasticity of neural networks
• Disentangling Transfer and Interference in Multi-Domain Learning
• Exploring the Challenges towards Lifelong Fact Learning
• TAG: Task-based Accumulated Gradients for Lifelong learning
• OvA-INN: Continual Learning with Invertible Neural Networks
• POET: Prompt Offset Tuning for Continual Human Action Adaptation
• Class Incremental Online Streaming Learning
• Regression Bugs Are In Your Model! Measuring, Reducing and Analyzing Regressions In NLP Model Updates
• Class-incremental Learning with Rectified Feature-Graph Preservation
• Online Continual Learning in Image Classification: An Empirical Survey
• Adaptive Explainable Continual Learning Framework for Regression Problems with Focus on Power Forecasts
• Shared and Private VAEs with Generative Replay for Continual Learning
• Kronecker Factorization for Preventing Catastrophic Forgetting in Large-scale Medical Entity Linking
• Target Layer Regularization for Continual Learning Using Cramer-Wold Generator
• Towards Continual Entity Learning in Language Models for Conversational Agents
• Deep Virtual Networks for Memory Efficient Inference of Multiple Tasks
• Compression-aware Continual Learning using Singular Value Decomposition
• Exploring the Tradeoff Between Diversity and Discrimination for Continuous Category Discovery