{"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/lstm-a-search-space-odyssey","title":"LSTM: A Search Space Odyssey","arxiv_id":"1503.04069","date":"2015-03-13","proceeding":null,"authors":["Klaus Greff","Rupesh Kumar Srivastava","Jan Koutník","Bas R. Steunebrink","Jürgen Schmidhuber"],"abstract":"Several variants of the Long Short-Term Memory (LSTM) architecture for\nrecurrent neural networks have been proposed since its inception in 1995. In\nrecent years, these networks have become the state-of-the-art models for a\nvariety of machine learning problems. This has led to a renewed interest in\nunderstanding the role and utility of various computational components of\ntypical LSTM variants. In this paper, we present the first large-scale analysis\nof eight LSTM variants on three representative tasks: speech recognition,\nhandwriting recognition, and polyphonic music modeling. The hyperparameters of\nall LSTM variants for each task were optimized separately using random search,\nand their importance was assessed using the powerful fANOVA framework. In\ntotal, we summarize the results of 5400 experimental runs ($\\approx 15$ years\nof CPU time), which makes our study the largest of its kind on LSTM networks.\nOur results show that none of the variants can improve upon the standard LSTM\narchitecture significantly, and demonstrate the forget gate and the output\nactivation function to be its most critical components. We further observe that\nthe studied hyperparameters are virtually independent and derive guidelines for\ntheir efficient adjustment.","url_abs":"http://arxiv.org/abs/1503.04069v2","url_pdf":"http://arxiv.org/pdf/1503.04069v2.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":"lstm-a-search-space-odyssey","repo_url":"https://github.com/RohanAsnani/capstone_project_machine_learning-nanodegree","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/a-martyn/unet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/alexandrujeman/dqn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/alexjeman/dqn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/disooqi/DNN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/disooqi/Natasy","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/flukeskywalker/highway-networks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/hadi-gharibi/pytorch-lstm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/kevin28520/RNN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/sw-scot/foco-mlmis","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/tmatha/lstm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/widmi/widis-lstm-tools","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/2023-MindSpore-1/ms-code-14/tree/main/LSTM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/2023-MindSpore-1/ms-code-219/tree/main/atae_lstm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/cui0523/Code6/tree/main/atae_lstm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/yangyucheng000/Paper-2/tree/main/lstm-gaf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"lstm-a-search-space-odyssey","repo_url":"https://github.com/yangyucheng000/papercode-2/tree/main/lstm-gaf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":null,"task_name":"CPU"},{"task_slug":"handwriting-recognition","task_name":"Handwriting Recognition"},{"task_slug":"music-modeling","task_name":"Music Modeling"},{"task_slug":"speech-recognition","task_name":"Speech Recognition"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1503.04069","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}