{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/neural-collapse-inspired-feature-classifier","title":"Neural Collapse Inspired Feature-Classifier Alignment for Few-Shot Class-Incremental Learning","arxiv_id":null,"date":"2023-02-02","proceeding":"ICLR 2023 2","authors":["Yibo Yang","Haobo Yuan","Xiangtai Li","Zhouchen Lin","Philip Torr","DaCheng Tao"],"abstract":"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. Finetuning the backbone or adjusting the classifier prototypes trained in the prior sessions would inevitably cause a misalignment between the feature and classifier of old classes, which explains the well-known catastrophic forgetting problem. In this paper, we deal with this misalignment dilemma in FSCIL inspired by the recently discovered phenomenon named neural collapse, which reveals that the last-layer features of the same class will collapse into a vertex, and the vertices of all classes are aligned with the classifier prototypes, which are formed as a simplex equiangular tight frame (ETF). It corresponds to an optimal geometric structure for classification due to the maximized Fisher Discriminant Ratio. We propose a neural collapse inspired framework for FSCIL. A group of classifier prototypes are pre-assigned as a simplex ETF for the whole label space, including the base session and all the incremental sessions. During training, the classifier prototypes are not learnable, and we adopt a novel loss function that drives the features into their corresponding prototypes. Theoretical analysis shows that our method holds the neural collapse optimality and does not break the feature-classifier alignment in an incremental fashion. Experiments on the miniImageNet, CUB-200, and CIFAR-100 datasets demonstrate that our proposed framework outperforms the state-of-the-art performances. Our code will be publicly available.","url_abs":"https://openreview.net/forum?id=y5W8tpojhtJ","url_pdf":"https://openreview.net/pdf?id=y5W8tpojhtJ","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"neural-collapse-inspired-feature-classifier","repo_url":"https://github.com/NeuralCollapseApplications/FSCIL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"class-incremental-learning","task_name":"Class Incremental Learning"},{"task_slug":"few-shot-class-incremental-learning","task_name":"Few-Shot Class-Incremental Learning"},{"task_slug":"incremental-learning","task_name":"Incremental Learning"},{"task_slug":"class-incremental-learning-1","task_name":"class-incremental learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/few-shot-class-incremental-learning-on-cifar","task":"Few-Shot Class-Incremental Learning","dataset":"CIFAR-100","model":"NC-FSCIL","rank_in_archive_order":5,"of":11,"metrics":{"Average Accuracy":"67.50","Last Accuracy":"56.11"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-class-incremental-learning-on-cub","task":"Few-Shot Class-Incremental Learning","dataset":"CUB-200-2011","model":"NC-FSCIL","rank_in_archive_order":6,"of":6,"metrics":{"Average Accuracy":"67.28","Last Accuracy ":"59.44"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-class-incremental-learning-on-mini","task":"Few-Shot Class-Incremental Learning","dataset":"mini-Imagenet","model":"NC-FSCIL","rank_in_archive_order":5,"of":12,"metrics":{"Average Accuracy":"67.82","Last Accuracy ":"58.31"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}