{"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-regularized-framework-for-sparse-and","title":"A Regularized Framework for Sparse and Structured Neural Attention","arxiv_id":"1705.07704","date":"2017-05-22","proceeding":"NeurIPS 2017 12","authors":["Vlad Niculae","Mathieu Blondel"],"abstract":"Modern neural networks are often augmented with an attention mechanism, which\ntells the network where to focus within the input. We propose in this paper a\nnew framework for sparse and structured attention, building upon a smoothed max\noperator. We show that the gradient of this operator defines a mapping from\nreal values to probabilities, suitable as an attention mechanism. Our framework\nincludes softmax and a slight generalization of the recently-proposed sparsemax\nas special cases. However, we also show how our framework can incorporate\nmodern structured penalties, resulting in more interpretable attention\nmechanisms, that focus on entire segments or groups of an input. We derive\nefficient algorithms to compute the forward and backward passes of our\nattention mechanisms, enabling their use in a neural network trained with\nbackpropagation. To showcase their potential as a drop-in replacement for\nexisting ones, we evaluate our attention mechanisms on three large-scale tasks:\ntextual entailment, machine translation, and sentence summarization. Our\nattention mechanisms improve interpretability without sacrificing performance;\nnotably, on textual entailment and summarization, we outperform the standard\nattention mechanisms based on softmax and sparsemax.","url_abs":"http://arxiv.org/abs/1705.07704v3","url_pdf":"http://arxiv.org/pdf/1705.07704v3.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-regularized-framework-for-sparse-and","repo_url":"https://github.com/dhruvdcoder/sparse-structured-attention","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"a-regularized-framework-for-sparse-and","repo_url":"https://github.com/vene/sparse-structured-attention","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"a-regularized-framework-for-sparse-and","repo_url":"https://github.com/weiwang2330/sparse-structured-attention","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"sentence-summarization","task_name":"Sentence Summarization"},{"task_slug":"text-summarization","task_name":"Text Summarization"},{"task_slug":"translation","task_name":"Translation"}],"methods":[{"method_slug":"interpretability","method_name":"Interpretability"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.07704","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}