# A Closer Look at Prototype Classifier for Few-shot Image Classification

arXiv: https://arxiv.org/abs/2110.05076 · HTML: https://mlsanity.com/paper/2110.05076/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- Builds on 29 of its 42 references that are in the map.
- First posted on arXiv 11 Oct 2021.

## Builds on (most cited first, 29 of 29)

- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [Prototypical Networks for Few-shot Learning](https://mlsanity.com/paper/1703.05175/) (2017)
- [Matching Networks for One Shot Learning](https://mlsanity.com/paper/1606.04080/) (2016)
- [Learning to Compare: Relation Network for Few-Shot Learning](https://mlsanity.com/paper/1711.06025/) (2017)
- [CosFace: Large Margin Cosine Loss for Deep Face Recognition](https://mlsanity.com/paper/1801.09414/) (2018)
- [SphereFace: Deep Hypersphere Embedding for Face Recognition](https://mlsanity.com/paper/1704.08063/) (2017)
- [A Closer Look at Few-shot Classification](https://mlsanity.com/paper/1904.04232/) (2019)
- [Meta-Learning with Differentiable Convex Optimization](https://mlsanity.com/paper/1904.03758/) (2019)
- [Meta-Learning for Semi-Supervised Few-Shot Classification](https://mlsanity.com/paper/1803.00676/) (2018)
- [Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need?](https://mlsanity.com/paper/2003.11539/) (2020)
- [Meta-learning with differentiable closed-form solvers](https://mlsanity.com/paper/1805.08136/) (2018)
- [Probabilistic Model-Agnostic Meta-Learning](https://mlsanity.com/paper/1806.02817/) (2018)
- [Few-Shot Learning via Learning the Representation, Provably](https://mlsanity.com/paper/2002.09434/) (2020)
- [SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning](https://mlsanity.com/paper/1911.04623/) (2019)
- [CrossTransformers: spatially-aware few-shot transfer](https://mlsanity.com/paper/2007.11498/) (2020)
- [Free Lunch for Few-shot Learning: Distribution Calibration](https://mlsanity.com/paper/2101.06395/) (2021)
- [Negative Margin Matters: Understanding Margin in Few-shot Classification](https://mlsanity.com/paper/2003.12060/) (2020)
- [Improved Few-Shot Visual Classification](https://mlsanity.com/paper/1912.03432/) (2019)
- [Infinite Mixture Prototypes for Few-Shot Learning](https://mlsanity.com/paper/1902.04552/) (2019)
- [TapNet: Neural Network Augmented with Task-Adaptive Projection for Few-Shot Learning](https://mlsanity.com/paper/1905.06549/) (2019)
- [Self-training for Few-shot Transfer Across Extreme Task Differences](https://mlsanity.com/paper/2010.07734/) (2020)
- [Multimodal Prototypical Networks for Few-shot Learning](https://mlsanity.com/paper/2011.08899/) (2020)
- [On Episodes, Prototypical Networks, and Few-shot Learning](https://mlsanity.com/paper/2012.09831/) (2020)
- [A Theoretical Analysis of the Number of Shots in Few-Shot Learning](https://mlsanity.com/paper/1909.11722/) (2019)
- [Centroid-based deep metric learning for speaker recognition](https://mlsanity.com/paper/1902.02375/) (2019)
- [A Two-Stage Approach to Few-Shot Learning for Image Recognition](https://mlsanity.com/paper/1912.04973/) (2019)
- [Theoretical bounds on estimation error for meta-learning](https://mlsanity.com/paper/2010.07140/) (2020)
- [Max-margin Class Imbalanced Learning with Gaussian Affinity](https://mlsanity.com/paper/1901.07711/) (2019)
- [Few-shot Image Recognition with Manifolds](https://mlsanity.com/paper/2010.12084/) (2020)

## Built on by (most cited first, 15 of 15)

- [Prototype-Based Image Prompting for Weakly Supervised Histopathological Image Segmentation](https://mlsanity.com/paper/2503.12068/) (2025)
- [Learning Clustering-based Prototypes for Compositional Zero-shot Learning](https://mlsanity.com/paper/2502.06501/) (2025)
- [LSFSL: Leveraging Shape Information in Few-shot Learning](https://mlsanity.com/paper/2304.06672/) (2023)
- [Practical Continual Forgetting for Pre-trained Vision Models](https://mlsanity.com/paper/2501.09705/) (2025)
- [FedSC: Federated Learning with Semantic-Aware Collaboration](https://mlsanity.com/paper/2506.21012/) (2025)
- [Hyperspherical Forward-Forward with Prototypical Representations](https://mlsanity.com/paper/2605.00082/) (2026)
- [Mutual Information Guided Optimal Transport for Unsupervised Visible-Infrared Person Re-identification](https://mlsanity.com/paper/2407.12758/) (2024)
- [Any-stepsize Gradient Descent for Separable Data under Fenchel-Young Losses](https://mlsanity.com/paper/2502.04889/) (2025)
- [Enhancing Zero-Shot Image Recognition in Vision-Language Models through Human-like Concept Guidance](https://mlsanity.com/paper/2503.15886/) (2025)
- [Active Transfer Prototypical Network: An Efficient Labeling Algorithm for Time-Series Data](https://mlsanity.com/paper/2209.14199/) (2022)
- [SCNode: Spatial and Contextual Coordinates for Graph Representation Learning](https://mlsanity.com/paper/2410.02158/) (2024)
- [Semantically Guided Dynamic Visual Prototype Refinement for Compositional Zero-Shot Learning](https://mlsanity.com/paper/2501.07114/) (2025)
- [Domain Borders Are There to Be Crossed With Federated Few-Shot Adaptation](https://mlsanity.com/paper/2507.10160/) (2025)
- [Color as the Impetus: Transforming Few-Shot Learner](https://mlsanity.com/paper/2507.22136/) (2025)
- [Mochi: Aligning Pre-training and Inference for Efficient Graph Foundation Models via Meta-Learning](https://mlsanity.com/paper/2604.22031/) (2026)

## Abstract (arXiv)

The prototypical network is a prototype classifier based on meta-learning and is widely used for few-shot learning because it classifies unseen examples by constructing class-specific prototypes without adjusting hyper-parameters during meta-testing. Interestingly, recent research has attracted a lot of attention, showing that training a new linear classifier, which does not use a meta-learning algorithm, performs comparably with the prototypical network. However, the training of a new linear classifier requires the retraining of the classifier every time a new class appears. In this paper, we analyze how a prototype classifier works equally well without training a new linear classifier or meta-learning. We experimentally find that directly using the feature vectors, which is extracted by using standard pre-trained models to construct a prototype classifier in meta-testing, does not perform as well as the prototypical network and training new linear classifiers on the feature vectors of pre-trained models. Thus, we derive a novel generalization bound for a prototypical classifier and show that the transformation of a feature vector can improve the performance of prototype classifiers. We experimentally investigate several normalization methods for minimizing the derived bound and find that the same performance can be obtained by using the L2 normalization and minimizing the ratio of the within-class variance to the between-class variance without training a new classifier or meta-learning.