{"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/an-empirical-study-of-spatial-attention","title":"An Empirical Study of Spatial Attention Mechanisms in Deep Networks","arxiv_id":"1904.05873","date":"2019-04-11","proceeding":"ICCV 2019 10","authors":["Xizhou Zhu","Dazhi Cheng","Zheng Zhang","Stephen Lin","Jifeng Dai"],"abstract":"Attention mechanisms have become a popular component in deep neural networks,\nyet there has been little examination of how different influencing factors and\nmethods for computing attention from these factors affect performance. Toward a\nbetter general understanding of attention mechanisms, we present an empirical\nstudy that ablates various spatial attention elements within a generalized\nattention formulation, encompassing the dominant Transformer attention as well\nas the prevalent deformable convolution and dynamic convolution modules.\nConducted on a variety of applications, the study yields significant findings\nabout spatial attention in deep networks, some of which run counter to\nconventional understanding. For example, we find that the query and key content\ncomparison in Transformer attention is negligible for self-attention, but vital\nfor encoder-decoder attention. A proper combination of deformable convolution\nwith key content only saliency achieves the best accuracy-efficiency tradeoff\nin self-attention. Our results suggest that there exists much room for\nimprovement in the design of attention mechanisms.","url_abs":"http://arxiv.org/abs/1904.05873v1","url_pdf":"http://arxiv.org/pdf/1904.05873v1.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":"an-empirical-study-of-spatial-attention","repo_url":"https://github.com/open-mmlab/mmdetection","is_official":0,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"deformable-convolution","method_name":"Deformable Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1904.05873","atlas_url":"https://app.syntology.ai/?focus=1904.05873","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}