{"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/learning-to-remember-rare-events","title":"Learning to Remember Rare Events","arxiv_id":"1703.03129","date":"2017-03-09","proceeding":null,"authors":["Łukasz Kaiser","Ofir Nachum","Aurko Roy","Samy Bengio"],"abstract":"Despite recent advances, memory-augmented deep neural networks are still\nlimited when it comes to life-long and one-shot learning, especially in\nremembering rare events. We present a large-scale life-long memory module for\nuse in deep learning. The module exploits fast nearest-neighbor algorithms for\nefficiency and thus scales to large memory sizes. Except for the\nnearest-neighbor query, the module is fully differentiable and trained\nend-to-end with no extra supervision. It operates in a life-long manner, i.e.,\nwithout the need to reset it during training.\n  Our memory module can be easily added to any part of a supervised neural\nnetwork. To show its versatility we add it to a number of networks, from simple\nconvolutional ones tested on image classification to deep sequence-to-sequence\nand recurrent-convolutional models. In all cases, the enhanced network gains\nthe ability to remember and do life-long one-shot learning. Our module\nremembers training examples shown many thousands of steps in the past and it\ncan successfully generalize from them. We set new state-of-the-art for one-shot\nlearning on the Omniglot dataset and demonstrate, for the first time, life-long\none-shot learning in recurrent neural networks on a large-scale machine\ntranslation task.","url_abs":"http://arxiv.org/abs/1703.03129v1","url_pdf":"http://arxiv.org/pdf/1703.03129v1.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":"learning-to-remember-rare-events","repo_url":"https://github.com/tensorflow/models","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"learning-to-remember-rare-events","repo_url":"https://github.com/rdspring1/lsh_deeplearning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"few-shot-image-classification","task_name":"Few-Shot Image Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"one-shot-learning","task_name":"One-Shot Learning"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/few-shot-image-classification-on-omniglot-1-1","task":"Few-Shot Image Classification","dataset":"OMNIGLOT - 1-Shot, 20-way","model":"ConvNet with Memory Module","rank_in_archive_order":14,"of":20,"metrics":{"Accuracy":"95%"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-image-classification-on-omniglot-1-2","task":"Few-Shot Image Classification","dataset":"OMNIGLOT - 1-Shot, 5-way","model":"ConvNet with Memory Module","rank_in_archive_order":13,"of":17,"metrics":{"Accuracy":"98.4"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-image-classification-on-omniglot-5-1","task":"Few-Shot Image Classification","dataset":"OMNIGLOT - 5-Shot, 20-way","model":"ConvNet with Memory Module","rank_in_archive_order":11,"of":19,"metrics":{"Accuracy":"98.6%"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-image-classification-on-omniglot-5-2","task":"Few-Shot Image Classification","dataset":"OMNIGLOT - 5-Shot, 5-way","model":"ConvNet with Memory Module","rank_in_archive_order":10,"of":16,"metrics":{"Accuracy":"99.6"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1703.03129","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}