Online Continual Learning with Dynamic Label Hierarchies

2026-05-12

AI Summary

This work addresses the challenges posed by dynamically evolving label hierarchies in online continual learning, including partial supervision, granularity-dependent interference, semantic inconsistency, and catastrophic forgetting. To tackle these issues, the paper introduces Dynamic Hierarchical Online Continual Learning (DHOCL), a novel setting, and proposes the HALO method. HALO integrates an adaptive multi-head classification architecture with structured, learnable hierarchical prototypes, leveraging hierarchical regularization and experience replay to preserve cross-granularity semantic consistency and consolidate previously acquired knowledge. Extensive experiments demonstrate that HALO significantly outperforms existing approaches across multiple benchmarks, achieving state-of-the-art performance in terms of hierarchical accuracy, error severity, and overall continual learning capability.

Abstract

Online Continual Learning (OCL) aims to learn from endless non\text{-}stationary data streams, yet most existing methods assume a flat label space and overlook the hierarchical organization of real\text{-}world concepts that evolves both horizontally (sibling classes) and vertically (coarse or fine categories). To better reflect this context, we introduce a new problem setting, DHOCL (Online Continual Learning from Dynamic Hierarchies), where taxonomies evolve across granularities and each sample provides supervision at a single hierarchical level. In this setting, we find two fundamental issues: (i) partial supervision under mixed granularities provides only point-wise signals over an evolving path-wise hierarchy, which constrains plasticity and undermines cross-level semantic consistency, and (ii) the dynamically evolving hierarchies induce granularity-dependent interference, destabilizing popular replay and regularization mechanisms and thereby exacerbating catastrophic forgetting. To tackle these issues, we propose HALO (Hierarchical Adaptive Learning with Organized Prototypes), which adaptively combines complementary classification heads, regularized by organized learnable hierarchical prototypes, enabling rapid adaptation, hierarchical consistency, and structured knowledge consolidation as the taxonomy evolves. Extensive experiments on multiple benchmarks demonstrate that HALO consistently outperforms existing methods across hierarchical accuracy, mistake severity, and continual performance.

Problem

• Online Continual Learning
• Dynamic Label Hierarchies
• Catastrophic Forgetting
• Hierarchical Classification
• Non-stationary Data Streams

Innovation

• Online Continual Learning
• Dynamic Label Hierarchies
• Hierarchical Prototypes
• Catastrophic Forgetting
• Granularity-aware Learning