{"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/efficient-lifelong-learning-with-a-gem","title":"Efficient Lifelong Learning with A-GEM","arxiv_id":"1812.00420","date":"2018-12-02","proceeding":"ICLR 2019 5","authors":["Arslan Chaudhry","Marc'Aurelio Ranzato","Marcus Rohrbach","Mohamed Elhoseiny"],"abstract":"In lifelong learning, the learner is presented with a sequence of tasks,\nincrementally building a data-driven prior which may be leveraged to speed up\nlearning of a new task. In this work, we investigate the efficiency of current\nlifelong approaches, in terms of sample complexity, computational and memory\ncost. Towards this end, we first introduce a new and a more realistic\nevaluation protocol, whereby learners observe each example only once and\nhyper-parameter selection is done on a small and disjoint set of tasks, which\nis not used for the actual learning experience and evaluation. Second, we\nintroduce a new metric measuring how quickly a learner acquires a new skill.\nThird, we propose an improved version of GEM (Lopez-Paz & Ranzato, 2017),\ndubbed Averaged GEM (A-GEM), which enjoys the same or even better performance\nas GEM, while being almost as computationally and memory efficient as EWC\n(Kirkpatrick et al., 2016) and other regularization-based methods. Finally, we\nshow that all algorithms including A-GEM can learn even more quickly if they\nare provided with task descriptors specifying the classification tasks under\nconsideration. Our experiments on several standard lifelong learning benchmarks\ndemonstrate that A-GEM has the best trade-off between accuracy and efficiency.","url_abs":"http://arxiv.org/abs/1812.00420v2","url_pdf":"http://arxiv.org/pdf/1812.00420v2.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":"efficient-lifelong-learning-with-a-gem","repo_url":"https://github.com/facebookresearch/agem","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"efficient-lifelong-learning-with-a-gem","repo_url":"https://github.com/ContinualAI/avalanche","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"efficient-lifelong-learning-with-a-gem","repo_url":"https://github.com/aimagelab/mammoth","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"class-incremental-learning","task_name":"Class Incremental Learning"},{"task_slug":"continual-learning","task_name":"Continual Learning"},{"task_slug":"lifelong-learning","task_name":"Lifelong learning"},{"task_slug":"class-incremental-learning-1","task_name":"class-incremental learning"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/class-incremental-learning-on-cifar100","task":"Class Incremental Learning","dataset":"cifar100","model":"A-GEM","rank_in_archive_order":7,"of":7,"metrics":{"10-stage average accuracy":"45.76"},"uses_additional_data":false},{"leaderboard":"/sota/continual-learning-on-asc-19-tasks","task":"Continual Learning","dataset":"ASC (19 tasks)","model":"A-GEM","rank_in_archive_order":6,"of":15,"metrics":{"F1 - macro":"0.7844"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1812.00420","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1812.00420"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/aimagelab/mammoth","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ContinualAI/avalanche","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/facebookresearch/agem","reach":null}],"summary":{"ran_fixture":1,"unverified":11},"by_repo_kind":{"official":{"samples":12,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"f5ffbe87b7ec7cb6","entry":"create_fc_layer","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f5ffbe87b7ec7cb6"}},{"code_sha256_prefix":"fc44c3b7614307dc","entry":"Model","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"fc44c3b7614307dc"}},{"code_sha256_prefix":"60a81bb1d1fb0b67","entry":"_bn","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"60a81bb1d1fb0b67"}},{"code_sha256_prefix":"0562f5e312d29cbd","entry":"_conv","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0562f5e312d29cbd"}},{"code_sha256_prefix":"3de3229cf0087e10","entry":"_fc","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3de3229cf0087e10"}},{"code_sha256_prefix":"6a7664c18c49e6ba","entry":"_residual_block","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"6a7664c18c49e6ba"}},{"code_sha256_prefix":"5cfb34725fa7d7cc","entry":"_residual_block_first","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5cfb34725fa7d7cc"}},{"code_sha256_prefix":"a96da1b71ad1a9ef","entry":"bias_variable","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a96da1b71ad1a9ef"}},{"code_sha256_prefix":"4531aa012b415174","entry":"create_conv_layer","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4531aa012b415174"}},{"code_sha256_prefix":"3dd896a5062484eb","entry":"vgg_conv_layer","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3dd896a5062484eb"}},{"code_sha256_prefix":"b1b9b69ee1f47548","entry":"vgg_fc_layer","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"b1b9b69ee1f47548"}},{"code_sha256_prefix":"70fae4aed55bdadc","entry":"weight_variable","repo":"facebookresearch/agem","repo_kind":"official","path":"model/model.py","file_url":"https://github.com/facebookresearch/agem/blob/HEAD/model/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"70fae4aed55bdadc"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}