Oriel — AI & technology worth noticing, every day.

Diagnosing Prototype Estimation in Few-Shot Class-Incremental Learning: Training-Free Geometry Selection and Unlabeled-Query Prototype Rectification

Paper · https://arxiv.org/abs/2611.00022

arXiv:2611.00022v1 Announce Type: new Abstract: Frozen visual encoders make prototype heads competitive in few-shot class-incremental learning (FSCIL), but it remains unclear how much of the residual head gap is sensitive to noisy novel-class means rather than to the subsequent changes introduced by classifier refitting. On identical features, we construct a retrospective matched counterfactual: first replace each five-shot novel mean with its complete-data counterpart under fixed cosine scoring, then fit a linear probe. Across sixteen strong encoder dataset pairs, mean replacement accounts for the majority of this ordered gap. Shot and classwise diagnostics support a finite-sample mechanism, whereas a weaker base-only encoder does not satisfy the same criterion. This diagnosis motivates ShrinkSel, a support-only selector over prototype radius and covariance shrinkage whose family contains shared LDA as an exact fallback. Across sixteen evaluated pairs, it improves LDA on fifteen by 0.04-1.52 points and selects the exact LDA endpoint on Aircraft. We further introduce Query Prototype Rectification (QPR), a batch-transductive extension that uses gated unlabeled queries to update novel directions while preserving the selected geometry. On CLIP ViT-B/16, QPR adds 1.21-2.03 points; exact query-wise self-exclusion retains the benefit, showing that the gains do not require self-inclusion. The update repairs scarce novel prototypes at a measurable base-class cost and weakens with plentiful supports or very small cohorts. Thus, the evidence available for prototype estimation is a central constraint in frozen-feature FSCIL.

First seen by Oriel on .

AI summary

The paper finds that noisy novel-class prototypes explain most of the gap between prototype heads and refitted linear probes in frozen-feature FSCIL, and proposes support-only geometry selection plus query-based rectification.

  • A matched counterfactual on identical features across sixteen strong encoder-dataset pairs attributes most of the ordered performance gap to replacing five-shot novel means with complete-data means.
  • ShrinkSel selects prototype radius and covariance shrinkage using support data, includes shared LDA as an exact fallback, and beats LDA on fifteen of sixteen pairs by 0.04–1.52 points.
  • On Aircraft, ShrinkSel selects the exact LDA endpoint.
  • Query Prototype Rectification (QPR) uses gated unlabeled query batches to update novel directions while preserving selected geometry; on CLIP ViT-B/16 it adds 1.21–2.03 points, including with query-wise self-exclusion.
  • QPR gains come with a measurable base-class cost and weaken when supports are plentiful or query cohorts are very small.

#few-shot learning #class-incremental learning #prototype estimation #transductive learning

Generated by gpt-5.6-sol. Check the original for details.

Aliases

SourceExternal IDURL
arXiv AI Papersoai:arXiv.org:2611.00022v1https://arxiv.org/abs/2611.00022 Read discussion