Deep Generative Dual Memory Network for Continual Learning

Abstract

Despite advances in deep learning, neural networks can only learn multiple tasks when trained on them jointly. When tasks arrive sequentially, they lose performance on previously learnt tasks. This phenomenon called catastrophic forgetting is a fundamental challenge to overcome before neural networks can learn continually from incoming data. In this work, we derive inspiration from human memory to develop an architecture capable of learning continuously from sequentially incoming tasks, while averting catastrophic forgetting. Specifically, our contributions are: (i) a dual memory architecture emulating the complementary learning systems (hippocampus and the neocortex) in the human brain, (ii) memory consolidation via generative replay of past experiences, (iii) demonstrating advantages of generative replay and dual memories via experiments, and (iv) improved performance retention on challenging tasks even for low capacity models. Our architecture displays many characteristics of the mammalian memory and provides insights on the connection between sleep and learning.

References in corpus (7)

• Conditional Generative Adversarial Nets
• Overcoming catastrophic forgetting in neural networks
• Continual Learning Through Synaptic Intelligence
• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• Gradient Episodic Memory for Continual Learning
• Continual Learning with Deep Generative Replay
• Training Gaussian Mixture Models at Scale via Coresets

Cited by in corpus (47)

• Three scenarios for continual learning
• Generative replay with feedback connections as a general strategy for continual learning
• Lifelong Generative Modeling
• Pseudo-Rehearsal: Achieving Deep Reinforcement Learning without Catastrophic Forgetting
• Domain-Incremental Continual Learning for Mitigating Bias in Facial Expression and Action Unit Recognition
• Overcoming Long-term Catastrophic Forgetting through Adversarial Neural Pruning and Synaptic Consolidation
• Continual Learning of a Mixed Sequence of Similar and Dissimilar Tasks
• Orthogonal Gradient Descent for Continual Learning
• MT-Opt: Continuous Multi-Task Robotic Reinforcement Learning at Scale
• An Investigation of Replay-based Approaches for Continual Learning
• Toward Understanding Catastrophic Forgetting in Continual Learning
• Continual Learning with Adaptive Weights (CLAW)
• An Adaptive Random Path Selection Approach for Incremental Learning
• Distilling Causal Effect of Data in Class-Incremental Learning
• Optimization and Generalization of Regularization-Based Continual Learning: a Loss Approximation Viewpoint
• Enhancing Consistency and Mitigating Bias: A Data Replay Approach for Incremental Learning
• Encoders and Ensembles for Task-Free Continual Learning
• Continual Learning with Knowledge Transfer for Sentiment Classification
• Gradient Episodic Memory with a Soft Constraint for Continual Learning
• Self-Supervised Learning Aided Class-Incremental Lifelong Learning
• Memory-Efficient Semi-Supervised Continual Learning: The World is its Own Replay Buffer
• Dynamic Node Embeddings from Edge Streams
• Automatic Recall Machines: Internal Replay, Continual Learning and the Brain
• Lifelong Neural Predictive Coding: Learning Cumulatively Online without Forgetting
• Continual Learning in Deep Neural Network by Using a Kalman Optimiser
• OpenLORIS-Object: A Robotic Vision Dataset and Benchmark for Lifelong Deep Learning
• Continual Learning Using Bayesian Neural Networks
• Continual Learning with Fully Probabilistic Models
• Continual Learning Using Multi-view Task Conditional Neural Networks
• Overcoming Catastrophic Forgetting by Neuron-level Plasticity Control
• Continual Learning Using World Models for Pseudo-Rehearsal
• Continual Learning: Tackling Catastrophic Forgetting in Deep Neural Networks with Replay Processes
• A Deep Learning Framework for Lifelong Machine Learning
• Lifelong Vehicle Trajectory Prediction Framework Based on Generative Replay
• Frosting Weights for Better Continual Training
• Incremental Embedding Learning via Zero-Shot Translation
• Incremental Concept Learning via Online Generative Memory Recall
• IROS 2019 Lifelong Robotic Vision Challenge -- Lifelong Object Recognition Report
• From Static to Dynamic Node Embeddings
• Interleaved Multitask Learning with Energy Modulated Learning Progress
• Unsupervised Class-Incremental Learning Through Confusion
• When Video Classification Meets Incremental Classes
• RSAC: Regularized Subspace Approximation Classifier for Lightweight Continuous Learning
• Continual learning using hash-routed convolutional neural networks
• LIRA: Lifelong Image Restoration from Unknown Blended Distortions
• Leveraging Semantics for Incremental Learning in Multi-Relational Embeddings
• GRIm-RePR: Prioritising Generating Important Features for Pseudo-Rehearsal