{"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/adding-attentiveness-to-the-neurons-in","title":"Adding Attentiveness to the Neurons in Recurrent Neural Networks","arxiv_id":"1807.04445","date":"2018-07-12","proceeding":"ECCV 2018 9","authors":["Pengfei Zhang","Jianru Xue","Cuiling Lan","Wen-Jun Zeng","Zhanning Gao","Nanning Zheng"],"abstract":"Recurrent neural networks (RNNs) are capable of modeling the temporal\ndynamics of complex sequential information. However, the structures of existing\nRNN neurons mainly focus on controlling the contributions of current and\nhistorical information but do not explore the different importance levels of\ndifferent elements in an input vector of a time slot. We propose adding a\nsimple yet effective Element-wiseAttention Gate (EleAttG) to an RNN block\n(e.g., all RNN neurons in a network layer) that empowers the RNN neurons to\nhave the attentiveness capability. For an RNN block, an EleAttG is added to\nadaptively modulate the input by assigning different levels of importance,\ni.e., attention, to each element/dimension of the input. We refer to an RNN\nblock equipped with an EleAttG as an EleAtt-RNN block. Specifically, the\nmodulation of the input is content adaptive and is performed at fine\ngranularity, being element-wise rather than input-wise. The proposed EleAttG,\nas an additional fundamental unit, is general and can be applied to any RNN\nstructures, e.g., standard RNN, Long Short-Term Memory (LSTM), or Gated\nRecurrent Unit (GRU). We demonstrate the effectiveness of the proposed\nEleAtt-RNN by applying it to the action recognition tasks on both 3D human\nskeleton data and RGB videos. Experiments show that adding attentiveness\nthrough EleAttGs to RNN blocks significantly boosts the power of RNNs.","url_abs":"http://arxiv.org/abs/1807.04445v1","url_pdf":"http://arxiv.org/pdf/1807.04445v1.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":[],"tasks":[{"task_slug":"action-recognition-in-videos","task_name":"Action Recognition"},{"task_slug":"skeleton-based-action-recognition","task_name":"Skeleton Based Action Recognition"},{"task_slug":"action-recognition","task_name":"Temporal Action Localization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/skeleton-based-action-recognition-on-ntu-rgbd","task":"Skeleton Based Action Recognition","dataset":"NTU RGB+D","model":"EleAtt-GRU","rank_in_archive_order":114,"of":135,"metrics":{"Accuracy (CS)":"79.8","Accuracy (CV)":"87.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1807.04445","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}