Neural Collapse by Design: Learning Class Prototypes on the Hypersphere

• URL: http://arxiv.org/abs/2605.20302v2
• Date: Thu, 21 May 2026 07:20:48 GMT
• Title: Neural Collapse by Design: Learning Class Prototypes on the Hypersphere
• Abstract summary: Cross entropy (CE) leaves radial degrees of freedom unconstrained and converges to a degenerate geometry, while supervised contrastive learning (SCL) drives features toward NC during pretraining but discards this structure in a post hoc linear probing phase.<n>We show that both paradigms are different appearances of the same method that contrasts prototypes on the unit hypersphere, and that closing the gap requires fixing each at its point of failure.<n>Our work recasts supervised learning as prototype learning on the hypersphere, with NC reached by design.
• Score: 19.046486593203323
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Supervised classification has a theoretical optimum, Neural Collapse (NC), yet neither of its two dominant paradigms reaches it in practice. Cross entropy (CE) leaves radial degrees of freedom unconstrained and converges to a degenerate geometry, while supervised contrastive learning (SCL) drives features toward NC during pretraining but discards this structure in a post hoc linear probing phase. We show that both paradigms are different appearances of the same method that contrasts prototypes on the unit hypersphere, and that closing the gap requires fixing each at its point of failure. From the CE side, we propose NTCE and NONL, two normalized losses that import contrastive optimization's missing ingredients into classifier learning: a large effective negative set and decoupled alignment and uniformity terms. From the SCL side, we prove that SCL's objective already optimizes throughout training for a principled classifier whose weights are the class mean embeddings, making linear probing both redundant and harmful. Empirically, on four benchmarks including ImageNet-1K, NTCE and NONL surpass CE accuracy, closely approximate NC ($\geq 95\%$), and match CE's converged NC on 4/5 metrics in under $7.5\%$ of its iterations, while SCL with fixed prototypes matches linear probing without the hours-long classifier training phase. The learned geometry yields $+5.5\%$ mean relative improvement in transfer learning, up to $+8.7\%$ under severe class imbalance, and improved robustness to corruptions on ImageNet-C. Our work recasts supervised learning as prototype learning on the hypersphere, with NC reached by design.

Related papers

• Cluster-Aware Neural Collapse Prompt Tuning for Long-Tailed Generalization of Vision-Language Models [20.68004586002534]
  We propose cluster-aware neural collapse prompt tuning (CPT)<n>CPT enhances the discriminability of tail classes in prompt-tuned vision-language models without sacrificing generalization.<n>Experiments on 11 diverse datasets demonstrate that CPT outperforms SOTA methods.
  arXiv  Detail & Related papers  (2026-05-12T10:50:43Z)
• Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning [48.11265601808718]
  We show that standard self-supervised contrastive learning objectives implicitly approximate a supervised variant we call the negatives-only supervised contrastive loss (NSCL)<n>We prove that the gap between the CL and NSCL losses vanishes as the number of semantic classes increases, under a bound that is both label-agnostic and architecture-independent.
  arXiv  Detail & Related papers  (2025-06-04T19:43:36Z)
• Class-Imbalanced Semi-Supervised Learning for Large-Scale Point Cloud Semantic Segmentation via Decoupling Optimization [64.36097398869774]
  Semi-supervised learning (SSL) has been an active research topic for large-scale 3D scene understanding. The existing SSL-based methods suffer from severe training bias due to class imbalance and long-tail distributions of the point cloud data. We introduce a new decoupling optimization framework, which disentangles feature representation learning and classifier in an alternative optimization manner to shift the bias decision boundary effectively.
  arXiv  Detail & Related papers  (2024-01-13T04:16:40Z)
• Neural Collapse for Cross-entropy Class-Imbalanced Learning with Unconstrained ReLU Feature Model [25.61363481391964]
  We show that when the training dataset is class-imbalanced, some Neural Collapse (NC) properties will no longer be true. In this paper, we generalize NC to imbalanced regime for cross-entropy loss under the unconstrained ReLU feature model. We find that the weights are aligned to the scaled and centered class-means with scaling factors depend on the number of training samples of each class.
  arXiv  Detail & Related papers  (2024-01-04T04:53:31Z)
• Symmetric Neural-Collapse Representations with Supervised Contrastive Loss: The Impact of ReLU and Batching [26.994954303270575]
  Supervised contrastive loss (SCL) is a competitive and often superior alternative to the cross-entropy loss for classification. While prior studies have demonstrated that both losses yield symmetric training representations under balanced data, this symmetry breaks under class imbalances. This paper presents an intriguing discovery: the introduction of a ReLU activation at the final layer effectively restores the symmetry in SCL-learned representations.
  arXiv  Detail & Related papers  (2023-06-13T17:55:39Z)
• Neural Collapse Inspired Feature-Classifier Alignment for Few-Shot Class Incremental Learning [120.53458753007851]
  Few-shot class-incremental learning (FSCIL) has been a challenging problem as only a few training samples are accessible for each novel class in the new sessions. We deal with this misalignment dilemma in FSCIL inspired by the recently discovered phenomenon named neural collapse. We propose a neural collapse inspired framework for FSCIL. Experiments on the miniImageNet, CUB-200, and CIFAR-100 datasets demonstrate that our proposed framework outperforms the state-of-the-art performances.
  arXiv  Detail & Related papers  (2023-02-06T18:39:40Z)
• Orthogonal SVD Covariance Conditioning and Latent Disentanglement [65.67315418971688]
  Inserting an SVD meta-layer into neural networks is prone to make the covariance ill-conditioned. We propose Nearest Orthogonal Gradient (NOG) and Optimal Learning Rate (OLR) Experiments on visual recognition demonstrate that our methods can simultaneously improve covariance conditioning and generalization.
  arXiv  Detail & Related papers  (2022-12-11T20:31:31Z)
• Do We Really Need a Learnable Classifier at the End of Deep Neural Network? [118.18554882199676]
  We study the potential of learning a neural network for classification with the classifier randomly as an ETF and fixed during training. Our experimental results show that our method is able to achieve similar performances on image classification for balanced datasets.
  arXiv  Detail & Related papers  (2022-03-17T04:34:28Z)
• Semi-supervised Contrastive Learning with Similarity Co-calibration [72.38187308270135]
  We propose a novel training strategy, termed as Semi-supervised Contrastive Learning (SsCL) SsCL combines the well-known contrastive loss in self-supervised learning with the cross entropy loss in semi-supervised learning. We show that SsCL produces more discriminative representation and is beneficial to few shot learning.
  arXiv  Detail & Related papers  (2021-05-16T09:13:56Z)
• Prevalence of Neural Collapse during the terminal phase of deep learning training [7.031848258307718]
  Modern practice for training classification deepnets involves a Terminal Phase of Training (TPT) During TPT, the training error stays effectively zero while training loss is pushed towards zero. The symmetric and very simple geometry induced by the TPT confers important benefits, including better performance, better generalization, and better interpretability.
  arXiv  Detail & Related papers  (2020-08-18T23:12:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.