{"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-clockwork-rnn","title":"A Clockwork RNN","arxiv_id":"1402.3511","date":"2014-02-14","proceeding":null,"authors":["Jan Koutník","Klaus Greff","Faustino Gomez","Jürgen Schmidhuber"],"abstract":"Sequence prediction and classification are ubiquitous and challenging\nproblems in machine learning that can require identifying complex dependencies\nbetween temporally distant inputs. Recurrent Neural Networks (RNNs) have the\nability, in theory, to cope with these temporal dependencies by virtue of the\nshort-term memory implemented by their recurrent (feedback) connections.\nHowever, in practice they are difficult to train successfully when the\nlong-term memory is required. This paper introduces a simple, yet powerful\nmodification to the standard RNN architecture, the Clockwork RNN (CW-RNN), in\nwhich the hidden layer is partitioned into separate modules, each processing\ninputs at its own temporal granularity, making computations only at its\nprescribed clock rate. Rather than making the standard RNN models more complex,\nCW-RNN reduces the number of RNN parameters, improves the performance\nsignificantly in the tasks tested, and speeds up the network evaluation. The\nnetwork is demonstrated in preliminary experiments involving two tasks: audio\nsignal generation and TIMIT spoken word classification, where it outperforms\nboth RNN and LSTM networks.","url_abs":"http://arxiv.org/abs/1402.3511v1","url_pdf":"http://arxiv.org/pdf/1402.3511v1.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-clockwork-rnn","repo_url":"https://github.com/LinRiver/Deep-reinforcement-Learning-on-Futures-Trading","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"a-clockwork-rnn","repo_url":"https://github.com/SylvanLiu/Exchange-Rate-Prediction-based-on-Brexit-Modelling","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"a-clockwork-rnn","repo_url":"https://github.com/SylvanLiu/ExchangeRatePrediction","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"a-clockwork-rnn","repo_url":"https://github.com/html1101/Science-Fair-2019-2020","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"a-clockwork-rnn","repo_url":"https://github.com/pannous/caffe-speech-recognition","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"caffe2","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"}],"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=1402.3511","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}