Incremental Learning with Repetition via Pseudo-Feature Projection

2025-02-27

AI Summary

Real-world continual learning involves natural class repetition, ambiguous task boundaries, and no exemplar replay—posing significant challenges for incremental learning. Method: We propose an exemplar-free paradigm that models intra-class repetition via pseudo-feature projection to enable dynamic cross-task feature alignment. Our core method, Horde, dynamically integrates self-sustaining feature extractors and leverages implicit class repetition to drive self-supervised alignment. Contribution/Results: We introduce the first incremental learning benchmark explicitly incorporating intrinsic class repetition, systematically revealing how repetition affects exemplar-free methods. Experiments demonstrate that Horde matches state-of-the-art performance on classical non-repetitive benchmarks while achieving new SOTA results on repetitive scenarios. Crucially, it significantly enhances model robustness and generalization to concept re-emergence—without requiring stored samples or explicit task boundaries.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Search and Optimization: Learning to Search
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Incremental Learning scenarios do not always represent real-world inference use-cases, which tend to have less strict task boundaries, and exhibit repetition of common classes and concepts in their continual data stream. To better represent these use-cases, new scenarios with partial repetition and mixing of tasks are proposed, where the repetition patterns are innate to the scenario and unknown to the strategy. We investigate how exemplar-free incremental learning strategies are affected by data repetition, and we adapt a series of state-of-the-art approaches to analyse and fairly compare them under both settings. Further, we also propose a novel method (Horde), able to dynamically adjust an ensemble of self-reliant feature extractors, and align them by exploiting class repetition. Our proposed exemplar-free method achieves competitive results in the classic scenario without repetition, and state-of-the-art performance in the one with repetition.

Problem

• Analyzes incremental learning with data repetition.
• Proposes new scenarios with partial task repetition.
• Introduces Horde method for dynamic feature alignment.

Innovation

• Dynamic ensemble adjustment
• Exploiting class repetition
• Pseudo-feature projection method