{"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/compressing-recurrent-neural-networks-with","title":"Compressing Recurrent Neural Networks with Tensor Ring for Action Recognition","arxiv_id":"1811.07503","date":"2018-11-19","proceeding":"NIPS Workshop CDNNRIA 2018","authors":["Yu Pan","Jing Xu","Maolin Wang","Jinmian Ye","Fei Wang","Kun Bai","Zenglin Xu"],"abstract":"Recurrent Neural Networks (RNNs) and their variants, such as Long-Short Term\nMemory (LSTM) networks, and Gated Recurrent Unit (GRU) networks, have achieved\npromising performance in sequential data modeling. The hidden layers in RNNs\ncan be regarded as the memory units, which are helpful in storing information\nin sequential contexts. However, when dealing with high dimensional input data,\nsuch as video and text, the input-to-hidden linear transformation in RNNs\nbrings high memory usage and huge computational cost. This makes the training\nof RNNs unscalable and difficult. To address this challenge, we propose a novel\ncompact LSTM model, named as TR-LSTM, by utilizing the low-rank tensor ring\ndecomposition (TRD) to reformulate the input-to-hidden transformation. Compared\nwith other tensor decomposition methods, TR-LSTM is more stable. In addition,\nTR-LSTM can complete an end-to-end training and also provide a fundamental\nbuilding block for RNNs in handling large input data. Experiments on real-world\naction recognition datasets have demonstrated the promising performance of the\nproposed TR-LSTM compared with the tensor train LSTM and other state-of-the-art\ncompetitors.","url_abs":"http://arxiv.org/abs/1811.07503v1","url_pdf":"http://arxiv.org/pdf/1811.07503v1.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":"compressing-recurrent-neural-networks-with","repo_url":"https://github.com/tnbar/tednet","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"action-recognition-in-videos","task_name":"Action Recognition"},{"task_slug":"action-recognition","task_name":"Temporal Action Localization"},{"task_slug":"tensor-decomposition","task_name":"Tensor Decomposition"}],"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=1811.07503","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}