Three scenarios for continual learning

Abstract

Standard artificial neural networks suffer from the well-known issue of catastrophic forgetting, making continual or lifelong learning difficult for machine learning. In recent years, numerous methods have been proposed for continual learning, but due to differences in evaluation protocols it is difficult to directly compare their performance. To enable more structured comparisons, we describe three continual learning scenarios based on whether at test time task identity is provided and--in case it is not--whether it must be inferred. Any sequence of well-defined tasks can be performed according to each scenario. Using the split and permuted MNIST task protocols, for each scenario we carry out an extensive comparison of recently proposed continual learning methods. We demonstrate substantial differences between the three scenarios in terms of difficulty and in terms of how efficient different methods are. In particular, when task identity must be inferred (i.e., class incremental learning), we find that regularization-based approaches (e.g., elastic weight consolidation) fail and that replaying representations of previous experiences seems required for solving this scenario.

Extended version of work presented at the NeurIPS Continual Learning workshop (2018); 18 pages, 5 figures, 6 tables. Related to arXiv:1809.10635

References in corpus (2)

• Distilling the Knowledge in a Neural Network
• PathNet: Evolution Channels Gradient Descent in Super Neural Networks

Cited by in corpus (42)

• Federated Continual Learning via Knowledge Fusion: A Survey
• Continual Learning in Sensor-based Human Activity Recognition: an Empirical Benchmark Analysis
• Lifelong Teacher-Student Network Learning
• Avoiding Catastrophe: Active Dendrites Enable Multi-Task Learning in Dynamic Environments
• Continual Learning of a Mixed Sequence of Similar and Dissimilar Tasks
• Maintaining Discrimination and Fairness in Class Incremental Learning
• Efficient Continual Learning with Modular Networks and Task-Driven Priors
• Meta-Consolidation for Continual Learning
• Continual Learning via Inter-Task Synaptic Mapping
• The Effectiveness of Memory Replay in Large Scale Continual Learning
• A Domain-Agnostic Approach for Characterization of Lifelong Learning Systems
• A First Look at Class Incremental Learning in Deep Learning Mobile Traffic Classification
• Generative Feature Replay For Class-Incremental Learning
• Imbalanced Continual Learning with Partitioning Reservoir Sampling
• Towards lifelong learning of Recurrent Neural Networks for control design
• Powerpropagation: A sparsity inducing weight reparameterisation
• Towards Fair Affective Robotics: Continual Learning for Mitigating Bias in Facial Expression and Action Unit Recognition
• Lifelong Learning using Eigentasks: Task Separation, Skill Acquisition, and Selective Transfer
• Continual Learning Using Multi-view Task Conditional Neural Networks
• Learning Invariant Representation for Continual Learning
• Graceful Degradation and Related Fields
• Layerwise Optimization by Gradient Decomposition for Continual Learning
• Look At Me, No Replay! SurpriseNet: Anomaly Detection Inspired Class Incremental Learning
• Task-agnostic Continual Learning with Hybrid Probabilistic Models
• CatNet: Class Incremental 3D ConvNets for Lifelong Egocentric Gesture Recognition
• Federated Reconnaissance: Efficient, Distributed, Class-Incremental Learning
• Incremental Learning via Rate Reduction
• Generative Feature Replay with Orthogonal Weight Modification for Continual Learning
• Energy Aligning for Biased Models
• Regularization-based Continual Learning for Fault Prediction in Lithium-Ion Batteries
• Better Knowledge Retention through Metric Learning
• Condensed Composite Memory Continual Learning
• OvA-INN: Continual Learning with Invertible Neural Networks
• Dendritic Self-Organizing Maps for Continual Learning
• Self-Promoted Prototype Refinement for Few-Shot Class-Incremental Learning
• Principal Gradient Direction and Confidence Reservoir Sampling for Continual Learning
• Lifelong Mixture of Variational Autoencoders
• Center Loss Regularization for Continual Learning
• Split-and-Bridge: Adaptable Class Incremental Learning within a Single Neural Network
• Lifelong Learning Without a Task Oracle
• Continual Competitive Memory: A Neural System for Online Task-Free Lifelong Learning
• Toward industrial use of continual learning : new metrics proposal for class incremental learning