{"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/order-matters-sequence-to-sequence-for-sets","title":"Order Matters: Sequence to sequence for sets","arxiv_id":"1511.06391","date":"2015-11-19","proceeding":null,"authors":["Oriol Vinyals","Samy Bengio","Manjunath Kudlur"],"abstract":"Sequences have become first class citizens in supervised learning thanks to\nthe resurgence of recurrent neural networks. Many complex tasks that require\nmapping from or to a sequence of observations can now be formulated with the\nsequence-to-sequence (seq2seq) framework which employs the chain rule to\nefficiently represent the joint probability of sequences. In many cases,\nhowever, variable sized inputs and/or outputs might not be naturally expressed\nas sequences. For instance, it is not clear how to input a set of numbers into\na model where the task is to sort them; similarly, we do not know how to\norganize outputs when they correspond to random variables and the task is to\nmodel their unknown joint probability. In this paper, we first show using\nvarious examples that the order in which we organize input and/or output data\nmatters significantly when learning an underlying model. We then discuss an\nextension of the seq2seq framework that goes beyond sequences and handles input\nsets in a principled way. In addition, we propose a loss which, by searching\nover possible orders during training, deals with the lack of structure of\noutput sets. We show empirical evidence of our claims regarding ordering, and\non the modifications to the seq2seq framework on benchmark language modeling\nand parsing tasks, as well as two artificial tasks -- sorting numbers and\nestimating the joint probability of unknown graphical models.","url_abs":"http://arxiv.org/abs/1511.06391v4","url_pdf":"http://arxiv.org/pdf/1511.06391v4.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":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/arunppsg/set2set","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/davidtangGT/MEGNET","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/materialsvirtuallab/megnet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/pyg-team/pytorch_geometric","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/teddykoker/mpnn-for-quantum-chem","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/threelittlemonkeys/pointer-network-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/MindSpore-paper-code-2/code400/tree/main/Roberta_Seq2Seq","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/dmlc/dgl/blob/master/examples/README.md","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"order-matters-sequence-to-sequence-for-sets","repo_url":"https://github.com/pyg-team/pytorch_geometric/blob/master/torch_geometric/nn/aggr/set2set.py","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"language-modeling","task_name":"Language Modeling"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"seq2seq","method_name":"Seq2Seq"},{"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=1511.06391","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}