Overcoming catastrophic forgetting in neural networks

DOI10.1073/PNAS.1611835114zbMATH Open1404.92015arXiv1612.00796OpenAlexW2560647685WikidataQ37737121 ScholiaQ37737121MaRDI QIDQ4646167FDOQ4646167

Authors: James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A. Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwińska, Demis Hassabis, Claudia Clopath, Dharshan Kumaran, Raia Hadsell 

Publication date: 11 January 2019

Published in: Proceedings of the National Academy of Sciences (Search for Journal in Brave)

Abstract: The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Neural networks are not, in general, capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks which they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on the MNIST hand written digit dataset and by learning several Atari 2600 games sequentially.

Full work available at URL: https://arxiv.org/abs/1612.00796

Recommendations

• Catastrophic forgetting in simple networks: an analysis of the pseudorehearsal solution
• scientific article; zbMATH DE number 1928798
• Adversarial Feature Alignment: Avoid Catastrophic Forgetting in Incremental Task Lifelong Learning
• A comprehensive study of class incremental learning algorithms for visual tasks
• Toward training recurrent neural networks for lifelong learning

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Neural networks for/in biological studies, artificial life and related topics (92B20)

Cited In (55)

• An algorithm for learning representations of models with scarce data
• Open-world continual learning: unifying novelty detection and continual learning
• The role of diversity and ensemble learning in credit card fraud detection
• Artificial neural variability for deep learning: on overfitting, noise memorization, and catastrophic forgetting
• Replay in deep learning: current approaches and missing biological elements
• Progressive learning: a deep learning framework for continual learning
• Adaptive infinite dropout for noisy and sparse data streams
• Model-Centric Data Manifold: The Data Through the Eyes of the Model
• Distributed Bayesian learning with stochastic natural gradient expectation propagation and the posterior server
• Hierarchically structured task-agnostic continual learning
• A neurosymbolic cognitive architecture framework for handling novelties in open worlds
• Stimulus-Driven and Spontaneous Dynamics in Excitatory-Inhibitory Recurrent Neural Networks for Sequence Representation
• Universal statistics of Fisher information in deep neural networks: mean field approach^*
• Task-agnostic continual learning using online variational Bayes with fixed-point updates
• Quantum continual learning of quantum data realizing knowledge backward transfer
• Learning deep optimizer for blind image deconvolution
• Class incremental learning with KL constraint and multi-strategy exemplar selection for classification based on MMFA model
• Sequential changepoint detection in neural networks with checkpoints
• Robust federated learning under statistical heterogeneity via hessian-weighted aggregation
• A minimum free energy model of motor learning
• Bayesian filtering with multiple internal models: toward a theory of social intelligence
• Dynamic Consolidation for Continual Learning
• Deep Reinforcement Learning: A State-of-the-Art Walkthrough
• Single circuit in V1 capable of switching contexts during movement using an inhibitory population as a switch
• Learning invariant features in modulatory networks through conflict and ambiguity
• Blessing of dimensionality at the edge and geometry of few-shot learning
• Exact learning dynamics of deep linear networks with prior knowledge
• The inverse variance-flatness relation in stochastic gradient descent is critical for finding flat minima
• Gated Orthogonal Recurrent Units: On Learning to Forget
• Continuous learning of spiking networks trained with local rules
• Toward training recurrent neural networks for lifelong learning
• Deep Bayesian unsupervised lifelong learning
• A comprehensive study of class incremental learning algorithms for visual tasks
• Drifting neuronal representations: bug or feature?
• A neurodynamic model of the interaction between color perception and color memory
• Accelerating algebraic multigrid methods via artificial neural networks
• Leveraging viscous Hamilton-Jacobi PDEs for uncertainty quantification in scientific machine learning
• One step back, two steps forward: interference and learning in recurrent neural networks
• Title not available (Why is that?)
• Lifelong deep learning-based control of robot manipulators
• An analytical theory of curriculum learning in teacher–student networks*
• Reliable extrapolation of deep neural operators informed by physics or sparse observations
• Reinforcement learning in sparse-reward environments with hindsight policy gradients
• A neural model of schemas and memory encoding
• Title not available (Why is that?)
• KS(conf): a light-weight test if a multiclass classifier operates outside of its specifications
• Adaptive learning of effective dynamics for online modeling of complex systems
• Exact learning dynamics of deep linear networks with prior knowledge
• Dynamic neural Turing machine with continuous and discrete addressing schemes
• Adversarial Feature Alignment: Avoid Catastrophic Forgetting in Incremental Task Lifelong Learning
• Bio-inspired, task-free continual learning through activity regularization
• Catastrophic forgetting in simple networks: an analysis of the pseudorehearsal solution
• Automated Deep Learning: Neural Architecture Search Is Not the End
• Accelerating actor-critic-based algorithms via pseudo-labels derived from prior knowledge
• A three-way decision approach for dynamically expandable networks

This page was built for publication: Overcoming catastrophic forgetting in neural networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4646167)