{"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/riemannian-walk-for-incremental-learning","title":"Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence","arxiv_id":"1801.10112","date":"2018-01-30","proceeding":"ECCV 2018 9","authors":["Arslan Chaudhry","Puneet K. Dokania","Thalaiyasingam Ajanthan","Philip H. S. Torr"],"abstract":"Incremental learning (IL) has received a lot of attention recently, however,\nthe literature lacks a precise problem definition, proper evaluation settings,\nand metrics tailored specifically for the IL problem. One of the main\nobjectives of this work is to fill these gaps so as to provide a common ground\nfor better understanding of IL. The main challenge for an IL algorithm is to\nupdate the classifier whilst preserving existing knowledge. We observe that, in\naddition to forgetting, a known issue while preserving knowledge, IL also\nsuffers from a problem we call intransigence, inability of a model to update\nits knowledge. We introduce two metrics to quantify forgetting and\nintransigence that allow us to understand, analyse, and gain better insights\ninto the behaviour of IL algorithms. We present RWalk, a generalization of\nEWC++ (our efficient version of EWC [Kirkpatrick2016EWC]) and Path Integral\n[Zenke2017Continual] with a theoretically grounded KL-divergence based\nperspective. We provide a thorough analysis of various IL algorithms on MNIST\nand CIFAR-100 datasets. In these experiments, RWalk obtains superior results in\nterms of accuracy, and also provides a better trade-off between forgetting and\nintransigence.","url_abs":"http://arxiv.org/abs/1801.10112v3","url_pdf":"http://arxiv.org/pdf/1801.10112v3.pdf","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":"riemannian-walk-for-incremental-learning","repo_url":"https://github.com/facebookresearch/agem","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"riemannian-walk-for-incremental-learning","repo_url":"https://github.com/ContinualAI/avalanche","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"incremental-learning","task_name":"Incremental Learning"}],"methods":[{"method_slug":"ewc","method_name":"EWC"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1801.10112","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}