FearNet: Brain-Inspired Model for Incremental Learning

Abstract

Incremental class learning involves sequentially learning classes in bursts of examples from the same class. This violates the assumptions that underlie methods for training standard deep neural networks, and will cause them to suffer from catastrophic forgetting. Arguably, the best method for incremental class learning is iCaRL, but it requires storing training examples for each class, making it challenging to scale. Here, we propose FearNet for incremental class learning. FearNet is a generative model that does not store previous examples, making it memory efficient. FearNet uses a brain-inspired dual-memory system in which new memories are consolidated from a network for recent memories inspired by the mammalian hippocampal complex to a network for long-term storage inspired by medial prefrontal cortex. Memory consolidation is inspired by mechanisms that occur during sleep. FearNet also uses a module inspired by the basolateral amygdala for determining which memory system to use for recall. FearNet achieves state-of-the-art performance at incremental class learning on image (CIFAR-100, CUB-200) and audio classification (AudioSet) benchmarks.

To appear in ICLR 2018

References in corpus (4)

• Overcoming catastrophic forgetting in neural networks
• YouTube-8M: A Large-Scale Video Classification Benchmark
• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• Neurogenesis Deep Learning

Cited by in corpus (75)

• A Comprehensive Study of Class Incremental Learning Algorithms for Visual Tasks
• Re-evaluating Continual Learning Scenarios: A Categorization and Case for Strong Baselines
• Generative replay with feedback connections as a general strategy for continual learning
• LAMOL: LAnguage MOdeling for Lifelong Language Learning
• Pseudo-Rehearsal: Achieving Deep Reinforcement Learning without Catastrophic Forgetting
• Memory Replay GANs: learning to generate images from new categories without forgetting
• Domain-Incremental Continual Learning for Mitigating Bias in Facial Expression and Action Unit Recognition
• Triple Memory Networks: a Brain-Inspired Method for Continual Learning
• Continual Learning of a Mixed Sequence of Similar and Dissimilar Tasks
• Pseudo-Recursal: Solving the Catastrophic Forgetting Problem in Deep Neural Networks
• Cognitively-Inspired Model for Incremental Learning Using a Few Examples
• Unified Probabilistic Deep Continual Learning through Generative Replay and Open Set Recognition
• Online Continual Learning on Sequences
• Incremental Few-Shot Learning with Attention Attractor Networks
• Class-incremental Learning via Deep Model Consolidation
• RODEO: Replay for Online Object Detection
• Continual Learning with Adaptive Weights (CLAW)
• Continual Learning with Node-Importance based Adaptive Group Sparse Regularization
• Incremental Deep Neural Network Learning using Classification Confidence Thresholding
• Meta Continual Learning
• EEC: Learning to Encode and Regenerate Images for Continual Learning
• Self-Supervised Training Enhances Online Continual Learning
• CL-CaGAN: Capsule differential adversarial continuous learning for cross-domain hyperspectral anomaly detection
• REMIND Your Neural Network to Prevent Catastrophic Forgetting
• Beneficial Perturbation Network for designing general adaptive artificial intelligence systems
• Wandering Within a World: Online Contextualized Few-Shot Learning
• The Next Big Thing(s) in Unsupervised Machine Learning: Five Lessons from Infant Learning
• An Adaptive Random Path Selection Approach for Incremental Learning
• Meta-Learning with Sparse Experience Replay for Lifelong Language Learning
• Continual Learning in Recurrent Neural Networks
• Continual Learning on the Edge with TensorFlow Lite
• Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis
• Distillation Techniques for Pseudo-rehearsal Based Incremental Learning
• Incremental Learning with Maximum Entropy Regularization: Rethinking Forgetting and Intransigence
• Generative Feature Replay For Class-Incremental Learning
• Training Networks in Null Space of Feature Covariance for Continual Learning
• Unified Question Generation with Continual Lifelong Learning
• Variational Prototype Replays for Continual Learning
• On the role of neurogenesis in overcoming catastrophic forgetting
• CoLLIE: Continual Learning of Language Grounding from Language-Image Embeddings
• Memory Efficient Experience Replay for Streaming Learning
• XtarNet: Learning to Extract Task-Adaptive Representation for Incremental Few-Shot Learning
• Memory-Efficient Semi-Supervised Continual Learning: The World is its Own Replay Buffer
• Bayesian Optimized Continual Learning with Attention Mechanism
• Automatic Recall Machines: Internal Replay, Continual Learning and the Brain
• DRILL: Dynamic Representations for Imbalanced Lifelong Learning
• Unifying Regularisation Methods for Continual Learning
• CBCL-PR: A Cognitively Inspired Model for Class-Incremental Learning in Robotics
• Do We Need Fully Connected Output Layers in Convolutional Networks?
• Memory Efficient Class-Incremental Learning for Image Classification
• Continual Learning: Tackling Catastrophic Forgetting in Deep Neural Networks with Replay Processes
• Frosting Weights for Better Continual Training
• Optimizing Reusable Knowledge for Continual Learning via Metalearning
• CatNet: Class Incremental 3D ConvNets for Lifelong Egocentric Gesture Recognition
• Graph-Based Continual Learning
• Understanding Catastrophic Forgetting and Remembering in Continual Learning with Optimal Relevance Mapping
• Closed-Loop Memory GAN for Continual Learning
• Accretionary Learning with Deep Neural Networks
• Beneficial perturbation network for continual learning
• Towards Class-incremental Object Detection with Nearest Mean of Exemplars
• Incremental Learning via Rate Reduction
• Distribution Networks for Open Set Learning
• Self-Net: Lifelong Learning via Continual Self-Modeling
• Unsupervised Continual Learning Via Pseudo Labels
• Dynamic Continual Learning: Harnessing Parameter Uncertainty for Improved Network Adaptation
• Selective Replay Enhances Learning in Online Continual Analogical Reasoning
• Continual learning using hash-routed convolutional neural networks
• A Biologically Plausible Audio-Visual Integration Model for Continual Learning
• Dynamic VAEs with Generative Replay for Continual Zero-shot Learning
• Class Incremental Online Streaming Learning
• Using Fictitious Class Representations to Boost Discriminative Zero-Shot Learners
• Shared and Private VAEs with Generative Replay for Continual Learning
• RSAC: Regularized Subspace Approximation Classifier for Lightweight Continuous Learning
• GRIm-RePR: Prioritising Generating Important Features for Pseudo-Rehearsal
• A Simple Illustration of Interleaved Learning using Kalman Filter for Linear Least Squares