Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IPEC: Test-Time Incremental Prototype Enhancement Classifier for Few-Shot Learning

Wenwen Liao, Hang Ruan, Jianbo Yu, Xiaofeng Yang, Qingchao Jiang, Xuefeng Yan

TL;DR

This work proposes a novel test-time method called Incremental Prototype Enhancement Classifier (IPEC), a test-time method that optimizes prototype estimation by leveraging information from previous query samples, and ground this approach in a Bayesian interpretation.

Abstract

Metric-based few-shot approaches have gained significant popularity due to their relatively straightforward implementation, high interpret ability, and computational efficiency. However, stemming from the batch-independence assumption during testing, which prevents the model from leveraging valuable knowledge accumulated from previous batches. To address these challenges, we propose a novel test-time method called Incremental Prototype Enhancement Classifier (IPEC), a test-time method that optimizes prototype estimation by leveraging information from previous query samples. IPEC maintains a dynamic auxiliary set by selectively incorporating query samples that are classified with high confidence. To ensure sample quality, we design a robust dual-filtering mechanism that assesses each query sample based on both global prediction confidence and local discriminative ability. By aggregating this auxiliary set with the support set in subsequent tasks, IPEC builds progressively more stable and representative prototypes, effectively reducing its reliance on the initial support set. We ground this approach in a Bayesian interpretation, conceptualizing the support set as a prior and the auxiliary set as a data-driven posterior, which in turn motivates the design of a practical "warm-up and test" two-stage inference protocol. Extensive empirical results validate the superior performance of our proposed method across multiple few-shot classification tasks.

IPEC: Test-Time Incremental Prototype Enhancement Classifier for Few-Shot Learning

TL;DR

This work proposes a novel test-time method called Incremental Prototype Enhancement Classifier (IPEC), a test-time method that optimizes prototype estimation by leveraging information from previous query samples, and ground this approach in a Bayesian interpretation.

Abstract

Metric-based few-shot approaches have gained significant popularity due to their relatively straightforward implementation, high interpret ability, and computational efficiency. However, stemming from the batch-independence assumption during testing, which prevents the model from leveraging valuable knowledge accumulated from previous batches. To address these challenges, we propose a novel test-time method called Incremental Prototype Enhancement Classifier (IPEC), a test-time method that optimizes prototype estimation by leveraging information from previous query samples. IPEC maintains a dynamic auxiliary set by selectively incorporating query samples that are classified with high confidence. To ensure sample quality, we design a robust dual-filtering mechanism that assesses each query sample based on both global prediction confidence and local discriminative ability. By aggregating this auxiliary set with the support set in subsequent tasks, IPEC builds progressively more stable and representative prototypes, effectively reducing its reliance on the initial support set. We ground this approach in a Bayesian interpretation, conceptualizing the support set as a prior and the auxiliary set as a data-driven posterior, which in turn motivates the design of a practical "warm-up and test" two-stage inference protocol. Extensive empirical results validate the superior performance of our proposed method across multiple few-shot classification tasks.
Paper Structure (36 sections, 22 equations, 8 figures, 6 tables, 1 algorithm)

This paper contains 36 sections, 22 equations, 8 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Problem Formulation
  4. Prototypical Networks
  5. Methodology
  6. Confidence-Based Auxiliary Set Construction
  7. Global Confidence via Entropy.
  8. Local Confidence via Logit Margin.
  9. Sample Selection Rule.
  10. IPEC Testing Procedure
  11. Auxiliary Samples Removal Strategy
  12. Bayesian Interpretations of IPEC Design Components
  13. IPEC Enables Continual Knowledge Accumulation
  14. Support Set as Prior Initialization
  15. Warm-up Phase Analysis
  16. ...and 21 more sections

Figures (8)

  • Figure 1: Comparison of Traditional Test and Ours test method.
  • Figure 2: Pipeline of our proposed method.
  • Figure 3: Impact of the Number of Shots on Accuracy. IPEC demonstrates robust performance that is less sensitive to the number of shots.
  • Figure 4: Analysis of IPEC's Dynamic Reliance on Support Shots. The figure illustrates how IPEC's accuracy and the auxiliary set's memory consumption evolve as support shots are gradually removed at different rates (one shot removed every k batches). (a) 1-shot setting (b) 5-shot setting
  • Figure 5: Warm-up Phase Effectiveness. Increasing the warm-up duration yields significant initial accuracy gains that later diminish, while memory cost grows steadily.
  • ...and 3 more figures