Revisiting Pool-based Prompt Learning for Few-shot Class-incremental Learning

TL;DR

This paper investigates the limitations of pool-based prompt learning in Few-Shot Class-Incremental Learning (FSCIL), identifies token-dimension saturation as a key issue, and proposes a novel spatial prompting method that improves performance on FSCIL benchmarks.

Contribution

The paper introduces LGSP-Prompt, a spatial prompting approach that overcomes token saturation issues, advancing prompt-based FSCIL methods with state-of-the-art results.

Findings

LGSP-Prompt outperforms existing methods on FSCIL benchmarks.

Spatial prompts effectively mitigate token-dimension saturation.

The approach enhances both knowledge retention and learning of new classes.

Abstract

Few-Shot Class-Incremental Learning (FSCIL) faces dual challenges of data scarcity and incremental learning in real-world scenarios. While pool-based prompting methods have demonstrated success in traditional incremental learning, their effectiveness in FSCIL settings remains unexplored. This paper presents the first study of current prompt pool methods in FSCIL tasks, revealing an unanticipated performance degradation in incremental sessions. Through comprehensive analysis, we identify that this phenomenon stems from token-dimension saturation: with limited data, excessive prompts compete for task-relevant information, leading to model overfitting. Based on this finding, we propose LGSP-Prompt (Local-Global Spatial Prompting), which innovatively shifts pool-based prompt learning from the token dimension to the spatial dimension. LGSP-Prompt generates spatial prompts by synergistically combining local spatial features and global frequency-domain representations to highlight key patterns in input images. We construct two spatial prompt pools enabling dynamic prompt selection to maintain acquired knowledge while effectively learning novel sessions. Extensive experiments demonstrate that our approach achieves state-of-the-art performance across multiple FSCIL benchmarks, showing significant advantages in both base knowledge preservation and incremental learning. Our implementation is available at https://github.com/Jywsuperman/LGSP.