Ambiguity-Guided Learnable Distribution Calibration for Semi-Supervised Few-Shot Class-Incremental Learning

2025-07-31

AI Summary

Existing semi-supervised few-shot class-incremental learning (Semi-FSCIL) methods assume unlabeled data contain only current novel classes, contradicting real-world scenarios where unlabeled samples are mixed across base and all incremental novel classes. Method: We propose Generalized Semi-FSCIL (GSemi-FSCIL), a more realistic setting, and introduce an ambiguity-guided learnable distribution calibration mechanism: it dynamically rectifies novel-class feature distribution shifts using base-class samples to mitigate inter-class confusion; jointly leverages scarce labeled data and heterogeneous unlabeled data to co-optimize feature distribution calibration, ambiguity modeling, and dynamic adaptation. Contribution/Results: Our approach achieves significant improvements over state-of-the-art methods on three benchmark datasets. It is the first work to empirically validate the feasibility and effectiveness of semi-supervised FSCIL under a setting substantially closer to practical deployment conditions.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Learning to Search
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Few-Shot Class-Incremental Learning (FSCIL) focuses on models learning new concepts from limited data while retaining knowledge of previous classes. Recently, many studies have started to leverage unlabeled samples to assist models in learning from few-shot samples, giving rise to the field of Semi-supervised Few-shot Class-Incremental Learning (Semi-FSCIL). However, these studies often assume that the source of unlabeled data is only confined to novel classes of the current session, which presents a narrow perspective and cannot align well with practical scenarios. To better reflect real-world scenarios, we redefine Semi-FSCIL as Generalized Semi-FSCIL (GSemi-FSCIL) by incorporating both base and all the ever-seen novel classes in the unlabeled set. This change in the composition of unlabeled samples poses a new challenge for existing methods, as they struggle to distinguish between unlabeled samples from base and novel classes. To address this issue, we propose an Ambiguity-guided Learnable Distribution Calibration (ALDC) strategy. ALDC dynamically uses abundant base samples to correct biased feature distributions for few-shot novel classes. Experiments on three benchmark datasets show that our method outperforms existing works, setting new state-of-the-art results.

Problem

• Redefining Semi-FSCIL to include base and novel classes in unlabeled data
• Addressing challenge of distinguishing unlabeled samples from base and novel classes
• Proposing ALDC to correct biased feature distributions for few-shot novel classes

Innovation

• Learnable distribution calibration for feature correction
• Dynamic use of base samples for bias correction
• Generalized Semi-FSCIL with broader unlabeled data scope