Memory-Efficient Semi-Supervised Continual Learning: The World is its Own Replay Buffer

2021/01/23 by James Smith, Jonathan Balloch, Smith, James +5 · 4 citations
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Human Pose and Action Recognition #Machine Learning (cs.LG) #Multimodal Machine Learning Applications #cs.CV #cs.LG

paper · pdf · doi:10.48550/arxiv.2101.09536

Accepted by the 2021 International Joint Conference on Neural Networks (IJCNN 2021)

openalex publication_date 2021/01/23 · openalex created_date 2021/02/01 · arxiv created 2021/05/06 · arxiv updated 2021/05/07 · openalex updated_date 2026/07/28

Abstract

Rehearsal is a critical component for class-incremental continual learning, yet it requires a substantial memory budget. Our work investigates whether we can significantly reduce this memory budget by leveraging unlabeled data from an agent's environment in a realistic and challenging continual learning paradigm. Specifically, we explore and formalize a novel semi-supervised continual learning (SSCL) setting, where labeled data is scarce yet non-i.i.d. unlabeled data from the agent's environment is plentiful. Importantly, data distributions in the SSCL setting are realistic and therefore reflect object class correlations between, and among, the labeled and unlabeled data distributions. We show that a strategy built on pseudo-labeling, consistency regularization, Out-of-Distribution (OoD) detection, and knowledge distillation reduces forgetting in this setting. Our approach, DistillMatch, increases performance over the state-of-the-art by no less than 8.7% average task accuracy and up to 54.5% average task accuracy in SSCL CIFAR-100 experiments. Moreover, we demonstrate that DistillMatch can save up to 0.23 stored images per processed unlabeled image compared to the next best method which only saves 0.08. Our results suggest that focusing on realistic correlated distributions is a significantly new perspective, which accentuates the importance of leveraging the world's structure as a continual learning strategy.

Citations

• Semi-Supervised Learning with Deep Generative Models
• Catastrophic Forgetting, Rehearsal and Pseudorehearsal
• Learning without Forgetting
• B-CNN: Branch Convolutional Neural Network for Hierarchical Classification
• Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networks
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
• Efficient Lifelong Learning with A-GEM
• Three scenarios for continual learning
• On Tiny Episodic Memories in Continual Learning
• Manifold Graph with Learned Prototypes for Semi-Supervised Image Classification
• Measuring Catastrophic Forgetting in Neural Networks
• Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis
• FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
• Generative Feature Replay For Class-Incremental Learning
• Self-Supervised Visual Feature Learning With Deep Neural Networks: A Survey
• Learning Multiple Layers of Features from Tiny Images

Cited by

• Rehearsal-free and Task-free Online Continual Learning With Contrastive Prompt
• Divide-and-Conquer for Enhancing Unlabeled Learning, Stability, and Plasticity in Semi-supervised Continual Learning
• CLA: Latent Alignment for Online Continual Self-Supervised Learning
• Self-Reinforcing Prototype Evolution with Dual-Knowledge Cooperation for Semi-Supervised Lifelong Person Re-Identification

Related

• Divide-and-Conquer for Enhancing Unlabeled Learning, Stability, and Plasticity in Semi-supervised Continual Learning
• A Closer Look at Rehearsal-Free Continual Learning
• Representational Continuity for Unsupervised Continual Learning
• Online Continual Learning under Extreme Memory Constraints
• Online Continual Learning on a Contaminated Data Stream with Blurry Task Boundaries