Reinforced Continual Learning

Abstract

Most artificial intelligence models have limiting ability to solve new tasks faster, without forgetting previously acquired knowledge. The recently emerging paradigm of continual learning aims to solve this issue, in which the model learns various tasks in a sequential fashion. In this work, a novel approach for continual learning is proposed, which searches for the best neural architecture for each coming task via sophisticatedly designed reinforcement learning strategies. We name it as Reinforced Continual Learning. Our method not only has good performance on preventing catastrophic forgetting but also fits new tasks well. The experiments on sequential classification tasks for variants of MNIST and CIFAR-100 datasets demonstrate that the proposed approach outperforms existing continual learning alternatives for deep networks.

References in corpus (5)

• Neural Architecture Search with Reinforcement Learning
• Continual Learning Through Synaptic Intelligence
• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• Neural Optimizer Search with Reinforcement Learning
• Lifelong Learning with Dynamically Expandable Networks

Cited by in corpus (46)

• A continual learning survey: Defying forgetting in classification tasks
• A Survey of Deep RL and IL for Autonomous Driving Policy Learning
• Adaptive Aggregation Networks for Class-Incremental Learning
• Incremental Learning for Semantic Segmentation of Large-Scale Remote Sensing Data
• Online Continual Learning with Maximally Interfered Retrieval
• Gradient based sample selection for online continual learning
• GraphSAIL: Graph Structure Aware Incremental Learning for Recommender Systems
• Recent Advances of Continual Learning in Computer Vision: An Overview
• Online Coreset Selection for Rehearsal-based Continual Learning
• Dynamically Expandable Graph Convolution for Streaming Recommendation
• Online Hyperparameter Optimization for Class-Incremental Learning
• Federated Continual Learning with Weighted Inter-client Transfer
• Continual Learning in Low-rank Orthogonal Subspaces
• Efficient Continual Learning with Modular Networks and Task-Driven Priors
• Flattening Sharpness for Dynamic Gradient Projection Memory Benefits Continual Learning
• Progressive Memory Banks for Incremental Domain Adaptation
• Scalable and Order-robust Continual Learning with Additive Parameter Decomposition
• Selective Amnesia: On Efficient, High-Fidelity and Blind Suppression of Backdoor Effects in Trojaned Machine Learning Models
• A Neural Span-Based Continual Named Entity Recognition Model
• ACE: Adapting to Changing Environments for Semantic Segmentation
• Online Fast Adaptation and Knowledge Accumulation: a New Approach to Continual Learning
• Hierarchical Indian Buffet Neural Networks for Bayesian Continual Learning
• Neural Architecture Search for Class-incremental Learning
• Continual Learning using a Bayesian Nonparametric Dictionary of Weight Factors
• Supermasks in Superposition
• All by Myself: Learning Individualized Competitive Behaviour with a Contrastive Reinforcement Learning optimization
• Continual Representation Learning for Biometric Identification
• Wide Neural Networks Forget Less Catastrophically
• Continual Learning in the Presence of Repetition
• A Deep Learning Framework for Lifelong Machine Learning
• CAM-GAN: Continual Adaptation Modules for Generative Adversarial Networks
• Memory Efficient Class-Incremental Learning for Image Classification
• Block Contextual MDPs for Continual Learning
• Single-Net Continual Learning with Progressive Segmented Training (PST)
• Frosting Weights for Better Continual Training
• AlterSGD: Finding Flat Minima for Continual Learning by Alternative Training
• A Conceptual Framework for Lifelong Learning
• Lifelong Learning with Searchable Extension Units
• Iterative Model-Based Reinforcement Learning Using Simulations in the Differentiable Neural Computer
• Continual Machine Reading Comprehension via Uncertainty-aware Fixed Memory and Adversarial Domain Adaptation
• Provable Lifelong Learning of Representations
• ZS-IL: Looking Back on Learned Experiences For Zero-Shot Incremental Learning
• Continual learning using hash-routed convolutional neural networks
• Context selectivity with dynamic availability enables lifelong continual learning
• FFNB: Forgetting-Free Neural Blocks for Deep Continual Visual Learning
• TAG: Task-based Accumulated Gradients for Lifelong learning