{"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/revisiting-distillation-and-incremental","title":"Revisiting Distillation and Incremental Classifier Learning","arxiv_id":"1807.02802","date":"2018-07-08","proceeding":null,"authors":["Khurram Javed","Faisal Shafait"],"abstract":"One of the key differences between the learning mechanism of humans and\nArtificial Neural Networks (ANNs) is the ability of humans to learn one task at\na time. ANNs, on the other hand, can only learn multiple tasks simultaneously.\nAny attempts at learning new tasks incrementally cause them to completely\nforget about previous tasks. This lack of ability to learn incrementally,\ncalled Catastrophic Forgetting, is considered a major hurdle in building a true\nAI system. In this paper, our goal is to isolate the truly effective existing\nideas for incremental learning from those that only work under certain\nconditions. To this end, we first thoroughly analyze the current state of the\nart (iCaRL) method for incremental learning and demonstrate that the good\nperformance of the system is not because of the reasons presented in the\nexisting literature. We conclude that the success of iCaRL is primarily due to\nknowledge distillation and recognize a key limitation of knowledge\ndistillation, i.e, it often leads to bias in classifiers. Finally, we propose a\ndynamic threshold moving algorithm that is able to successfully remove this\nbias. We demonstrate the effectiveness of our algorithm on CIFAR100 and MNIST\ndatasets showing near-optimal results. Our implementation is available at\nhttps://github.com/Khurramjaved96/incremental-learning.","url_abs":"http://arxiv.org/abs/1807.02802v2","url_pdf":"http://arxiv.org/pdf/1807.02802v2.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":"revisiting-distillation-and-incremental","repo_url":"https://github.com/Khurramjaved96/incremental-learning","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"revisiting-distillation-and-incremental","repo_url":"https://github.com/einavyog/my-incremental-learning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"incremental-learning","task_name":"Incremental Learning"},{"task_slug":"knowledge-distillation","task_name":"Knowledge Distillation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1807.02802","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}