{"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/a-strategy-for-an-uncompromising-incremental","title":"A Strategy for an Uncompromising Incremental Learner","arxiv_id":"1705.00744","date":"2017-05-02","proceeding":null,"authors":["Ragav Venkatesan","Hemanth Venkateswara","Sethuraman Panchanathan","Baoxin Li"],"abstract":"Multi-class supervised learning systems require the knowledge of the entire\nrange of labels they predict. Often when learnt incrementally, they suffer from\ncatastrophic forgetting. To avoid this, generous leeways have to be made to the\nphilosophy of incremental learning that either forces a part of the machine to\nnot learn, or to retrain the machine again with a selection of the historic\ndata. While these hacks work to various degrees, they do not adhere to the\nspirit of incremental learning. In this article, we redefine incremental\nlearning with stringent conditions that do not allow for any undesirable\nrelaxations and assumptions. We design a strategy involving generative models\nand the distillation of dark knowledge as a means of hallucinating data along\nwith appropriate targets from past distributions. We call this technique,\nphantom sampling.We show that phantom sampling helps avoid catastrophic\nforgetting during incremental learning. Using an implementation based on deep\nneural networks, we demonstrate that phantom sampling dramatically avoids\ncatastrophic forgetting. We apply these strategies to competitive multi-class\nincremental learning of deep neural networks. Using various benchmark datasets\nand through our strategy, we demonstrate that strict incremental learning could\nbe achieved. We further put our strategy to test on challenging cases,\nincluding cross-domain increments and incrementing on a novel label space. We\nalso propose a trivial extension to unbounded-continual learning and identify\npotential for future development.","url_abs":"http://arxiv.org/abs/1705.00744v2","url_pdf":"http://arxiv.org/pdf/1705.00744v2.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":"a-strategy-for-an-uncompromising-incremental","repo_url":"https://github.com/ragavvenkatesan/Incremental-GAN","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"class-incremental-learning","task_name":"Class Incremental Learning"},{"task_slug":"continual-learning","task_name":"Continual Learning"},{"task_slug":"incremental-learning","task_name":"Incremental Learning"},{"task_slug":"philosophy","task_name":"Philosophy"},{"task_slug":"class-incremental-learning-1","task_name":"class-incremental learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}