{"type":"ScholarlyArticle","title":"Randomized Greedy Search for Structured Prediction: Amortized Inference and Learning","authors":[{"name":"Chao Ma","slug":"chao-ma","given":"Chao","family":"Ma","affiliation":"School of EECS , Oregon State University"},{"name":"F A Rezaur Rahman Chowdhury","slug":"f-a-rezaur-rahman-chowdhury","given":"F A Rezaur Rahman","family":"Chowdhury","affiliation":"School of EECS , Washington State University"},{"name":"Aryan Deshwal","slug":"aryan-deshwal","given":"Aryan","family":"Deshwal","affiliation":"School of EECS , Washington State University"},{"name":"Md Rakibul Islam","slug":"md-rakibul-islam","given":"Md Rakibul","family":"Islam","affiliation":"School of EECS , Washington State University"},{"name":"Janardhan Rao Doppa","slug":"janardhan-rao-doppa","given":"Janardhan Rao","family":"Doppa","affiliation":"School of EECS , Washington State University"},{"name":"Dan Roth","slug":"dan-roth","given":"Dan","family":"Roth","affiliation":"Department of Computer and Information Science , University of Pennsylvania"}],"abstract":"In a structured prediction problem, we need to learn a predictor that can produce a structured output given a structured input (e.g., part-of-speech tagging). The key learning and inference challenge is due to the exponential size of the structured output space. This paper makes four contributions towards the goal of a computationally-efficient inference and training approach for structured prediction that allows to employ complex models and to optimize for non-decomposable loss functions. First, we define a simple class of randomized greedy search (RGS) based inference procedures that leverage classification algorithms for simple outputs. Second, we develop a RGS specific learning approach for amortized inference that can quickly produce high-quality outputs for a given set of structured inputs. Third, we plug our amortized RGS inference solver inside the inner loop of parameter-learning algorithms (e.g., structured SVM) to improve the speed of training. Fourth, we perform extensive experiments on diverse structured prediction tasks. Results show that our proposed approach is competitive or better than many state-of-the-art approaches in spite of its simplicity.","doi":"10.24963/ijcai.2019/713","url":"https://ijcai.org/proceedings/2019/713","pdf":"https://ijcai.org/proceedings/2019/0713.pdf","markdown":"https://ijcai.org/proceedings/2019/713.md","bibtex":"https://ijcai.org/proceedings/2019/bibtex/713","volume":{"slug":"2019","title":"Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2019","url":"https://ijcai.org/proceedings/2019/","year":2019,"location":"Macao","dates":"10-16 August 2019","editors":"Sarit Kraus","isbn":"978-0-9992411-4-1","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2019"},"section":"Natural Language Processing","track":"Main track","number":713,"pages":{"first":5130,"last":5138},"keywords":["Natural Language Processing: Natural Language Processing","Machine Learning: Structured Prediction"],"videos":[],"references":[{"position":1,"raw":"[Belanger et al., 2017] David Belanger, Bishan Yang, and Andrew McCallum. End-to-end learning for structured pre- diction energy networks. In ICML, 2017.","title":"End-to-end learning for structured prediction energy networks","authors":"David Belanger, Bishan Yang, Andrew Mccallum","year":"2017","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Boyan and Moore, 2001] Justin Boyan and Andrew W. Moore. Learning evaluation functions to improve optimiza- tion by local search. JMLR, 2001.","title":"Learning evaluation functions to improve optimization by local search","authors":"Justin Boyan, Andrew W Moore","year":"2001","venue":"JMLR","doi":null,"url":null,"internal_url":null},{"position":3,"raw":"[Chang et al., 2015a] Kai-Wei Chang, Akshay Krishna- murthy, Alekh Agarwal, Hal Daumé III, and John Langford. Learning to search better than your teacher. In ICML, 2015.","title":"Learning to search better than your teacher","authors":"Chang","year":"2015","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Chang et al., 2015b] Kai-Wei Chang, Shyam Upadhyay, Gourab Kundu, and Dan Roth. Structural learning with amortized inference. In AAAI, 2015.","title":"Structural learning with amortized inference","authors":"Chang","year":"2015","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":5,"raw":"[Collins, 2002] Michael Collins. Ranking algorithms for named entity extraction: Boosting and the voted percep- tron. In ACL, 2002.","title":"Ranking algorithms for named entity extraction: Boosting and the voted perceptron","authors":"Michael Collins","year":"2002","venue":"ACL","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Daumé and Marcu, 2005] Hal Daumé, III and Daniel Marcu. Learning as search optimization: Approximate large mar- gin methods for structured prediction. In ICML, 2005.","title":"Daniel Marcu. Learning as search optimization: Approximate large margin methods for structured prediction","authors":"Marcu ; Hal Daumé, Daumé, Iii","year":"2005","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[Daumé et al., 2009] Hal Daumé, III, John Langford, and Daniel Marcu. Search-based structured prediction. MLJ, 2009.","title":"Search-based structured prediction","authors":"Daumé","year":"2009","venue":null,"doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Doppa et al., 2014a] Janardhan Rao Doppa, Alan Fern, and Prasad Tadepalli. HC-Search: A learning framework for search-based structured prediction. JAIR, 2014.","title":"HC-Search: A learning framework for search-based structured prediction","authors":"Doppa","year":"2014","venue":"JAIR","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"[Doppa et al., 2014b] Janardhan Rao Doppa, Alan Fern, and Prasad Tadepalli. Structured prediction via output space search. JMLR, 2014.","title":"Structured prediction via output space search","authors":"Doppa","year":"2014","venue":"JMLR","doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Doppa et al., 2014c] Janardhan Rao Doppa, Jun Yu, Chao Ma, Alan Fern, and Prasad Tadepalli. HC-Search for multi- label prediction: An empirical study. In AAAI, 2014.","title":"HC-Search for multilabel prediction: An empirical study","authors":"Doppa","year":"2014","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[Durrett and Klein, 2014] Greg Durrett and Dan Klein. A joint model for entity analysis: Coreference, typing, and linking. In TACL, 2014.","title":"A joint model for entity analysis: Coreference, typing, and linking","authors":"Klein Durrett, Greg Durrett, Dan Klein","year":"2014","venue":"TACL","doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Fern, 2010] Alan Fern. Speedup learning. In Encyclopedia of Machine Learning, pages 907-911. 2010.","title":"Speedup learning","authors":"Alan Fern","year":"2010","venue":"Encyclopedia of Machine Learning","doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Finley and Joachims, 2008] Thomas Finley and Thorsten Joachims. Training structural SVMs when exact inference is intractable. In ICML, 2008.","title":"Training structural SVMs when exact inference is intractable","authors":"Joachims Finley, Thomas Finley, Thorsten Joachims","year":"2008","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[Gygli et al., 2017] Michael Gygli, Mohammad Norouzi, and Anelia Angelova. Deep value networks learn to evaluate and iteratively refine structured outputs. In ICML, 2017.","title":"Deep value networks learn to evaluate and iteratively refine structured outputs","authors":"Michael Gygli, Mohammad Norouzi, Anelia Angelova","year":"2017","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":15,"raw":"[He et al., 2014] He He, Hal Daumé III, and Jason Eisner. Learning to search in branch and bound algorithms. In NIPS, 2014.","title":"Learning to search in branch and bound algorithms","authors":"He","year":"2014","venue":"NIPS","doi":null,"url":null,"internal_url":null},{"position":16,"raw":"[Honorio and Jaakkola, 2016] Jean Honorio and Tommi S. Jaakkola. Structured prediction: From gaussian perturba- tions to linear-time principled algorithms. In UAI, 2016.","title":"Structured prediction: From gaussian perturbations to linear-time principled algorithms","authors":"Jaakkola Honorio, Jean Honorio, Tommi S Jaakkola","year":"2016","venue":"UAI","doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[Huang et al., 2012] Liang Huang, Suphan Fayong, and Yang Guo. Structured perceptron with inexact search. In NACCL, 2012.","title":"Structured perceptron with inexact search","authors":"Liang Huang, Suphan Fayong, Yang Guo","year":"2012","venue":"NACCL","doi":null,"url":null,"internal_url":null},{"position":18,"raw":"[Kundu et al., 2013] Gourab Kundu, Vivek Srikumar, and Dan Roth. Margin-based decomposed amortized inference. In ACL, 2013.","title":"Margin-based decomposed amortized inference","authors":"Gourab Kundu, Vivek Srikumar, Dan Roth","year":"2013","venue":"ACL","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Lafferty et al., 2001] John Lafferty, Andrew McCallum, and Fernando Pereira. Conditional random fields: Probabilis- tic models for segmenting and labeling sequence data. In ICML, 2001.","title":"Conditional random fields: Probabilistic models for segmenting and labeling sequence data","authors":"Lafferty","year":"2001","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[Lam et al., 2015] Michael Lam, Janardhan Rao Doppa, Sin- isa Todorovic, and Thomas G. Dietterich. HC-search for structured prediction in computer vision. In CVPR, 2015.","title":"HC-search for structured prediction in computer vision","authors":"Michael Lam, Janardhan Rao Doppa, Sinisa Todorovic, Thomas G Dietterich","year":"2015","venue":"CVPR","doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Liu et al., 2015] Fayao Liu, Guosheng Lin, and Chunhua Shen. CRF learning with CNN features for image segmen- tation. Pattern Recognition, 2015.","title":"CRF learning with CNN features for image segmentation","authors":"Liu","year":"2015","venue":"Pattern Recognition","doi":null,"url":null,"internal_url":null},{"position":22,"raw":"[Lucchi et al., 2011] Aurélien Lucchi, Yunpeng Li, Xavier Boix Bosch, Kevin Smith, and Pascal Fua. Are spatial and global constraints really necessary for segmentation. In ICCV, 2011.","title":"Are spatial and global constraints really necessary for segmentation","authors":"Aurélien Lucchi, Yunpeng Li, Xavier Boix Bosch, Kevin Smith, Pascal Fua","year":"2011","venue":"ICCV","doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Ma et al., 2014] Chao Ma, Janardhan Rao Doppa, John Walker Orr, Prashanth Mannem, Xiaoli Z. Fern, Thomas G. Dietterich, and Prasad Tadepalli. Prune-and- score: Learning for greedy coreference resolution. In EMNLP, 2014.","title":"Prune-andscore: Learning for greedy coreference resolution","authors":"Chao Ma, Janardhan Rao Doppa, John Walker Orr, Prashanth Mannem, Xiaoli Z Fern, Thomas G Dietterich, Prasad Tadepalli","year":"2014","venue":"EMNLP","doi":null,"url":null,"internal_url":null},{"position":24,"raw":"[Ma et al., 2017] Chao Ma, Janardhan Rao Doppa, Prasad Tadepalli, Hamed Shahbazi, and Xiaoli Z. Fern. Multi- task structured prediction for entity analysis: Search-based learning algorithms. In ACML, 2017.","title":"Multitask structured prediction for entity analysis: Search-based learning algorithms","authors":"Chao Ma, Janardhan Rao Doppa, Prasad Tadepalli, Hamed Shahbazi, Xiaoli Z Fern","year":"2017","venue":"ACML","doi":null,"url":null,"internal_url":null},{"position":25,"raw":"[Namaki et al., 2017] Mohammad Hossein Namaki, F. A. Rezaur Rahman Chowdhury, Md Rakibul Islam, Janard- han Rao Doppa, and Yinghui Wu. Learning to speed up query planning in graph databases. In ICAPS, 2017.","title":"Learning to speed up query planning in graph databases","authors":"Mohammad Hossein Namaki, F A Rezaur Rahman Chowdhury, Md Rakibul Islam, Janardhan Rao Doppa, Yinghui Wu","year":"2017","venue":"ICAPS","doi":null,"url":null,"internal_url":null},{"position":26,"raw":"[Pan and Srikumar, 2018] Xingyuan Pan and Vivek Sriku- mar. Learning to speed up structured output prediction. In ICML, 2018.","title":"Xingyuan Pan and Vivek Srikumar. Learning to speed up structured output prediction","authors":"Srikumar Pan","year":"2018","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":27,"raw":"[Ross et al., 2011] Stéphane Ross, Geoffery Gordon, and Drew Bagnell. A reduction of imitation learning and struc- tured prediction to no-regret online learning. In AISTATS, 2011.","title":"A reduction of imitation learning and structured prediction to no-regret online learning","authors":"Stéphane Ross, Geoffery Gordon, Drew Bagnell","year":"2011","venue":"AISTATS","doi":null,"url":null,"internal_url":null},{"position":28,"raw":"[Srikumar et al., 2012] Vivek Srikumar, Gourab Kundu, and Dan Roth. On amortizing inference cost for structured pre- diction. In EMNLP-CoNLL, 2012.","title":"On amortizing inference cost for structured prediction","authors":"Vivek Srikumar, Gourab Kundu, Dan Roth","year":"2012","venue":"EMNLP-CoNLL","doi":null,"url":null,"internal_url":null},{"position":29,"raw":"[Tsochantaridis et al., 2004] Ioannis Tsochantaridis, Thomas Hofmann, Thorsten Joachims, and Yasemin Altun. Support vector machine learning for interdependent and structured output spaces. In ICML, 2004.","title":"Support vector machine learning for interdependent and structured output spaces","authors":"Ioannis Tsochantaridis, Thomas Hofmann, Thorsten Joachims, Yasemin Altun","year":"2004","venue":"ICML","doi":null,"url":null,"internal_url":null},{"position":30,"raw":"[Tu and Gimpel, 2018] Lifu Tu and Kevin Gimpel. Learning approximate inference networks for structured prediction. In ICLR, 2018.","title":"Learning approximate inference networks for structured prediction","authors":"Gimpel ; Tu, Lifu Tu, Kevin Gimpel","year":"2018","venue":"ICLR","doi":null,"url":null,"internal_url":null},{"position":31,"raw":"[Weiss and Taskar, 2010] David Weiss and Ben Taskar. Structured prediction cascades. In AISTATS, 2010.","title":"Structured prediction cascades","authors":"David Weiss, Ben Taskar","year":"2010","venue":"AISTATS","doi":null,"url":null,"internal_url":null},{"position":32,"raw":"[Wiseman and Rush, 2016] Sam Wiseman and Alexander M. Rush. Sequence-to-sequence learning as beam-search op- timization. In EMNLP, 2016.","title":"Sequence-to-sequence learning as beam-search optimization","authors":"Sam Wiseman, Alexander M Rush","year":"2016","venue":"EMNLP","doi":null,"url":null,"internal_url":null},{"position":33,"raw":"[Xie et al., 2015] Jun Xie, Chao Ma, Janardhan Rao Doppa, Prashanth Mannem, Xiaoli Z. Fern, Thomas G. Dietterich, and Prasad Tadepalli. Learning greedy policies for the easy-first framework. In AAAI, 2015.","title":"Learning greedy policies for the easy-first framework","authors":"Jun Xie, Chao Ma, Janardhan Rao Doppa, Prashanth Mannem, Xiaoli Z Fern, Thomas G Dietterich, Prasad Tadepalli","year":"2015","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":34,"raw":"[Xu et al., 2009] Yuehua Xu, Alan Fern, and Sung Wook Yoon. Learning linear ranking functions for beam search with application to planning. JMLR, 10:1571-1610, 2009.","title":"Learning linear ranking functions for beam search with application to planning","authors":"Yuehua Xu, Alan Fern, Sung Wook Yoon","year":"2009","venue":"JMLR","doi":null,"url":null,"internal_url":null},{"position":35,"raw":"[Zhang et al., 2014] Yuan Zhang, Tao Lei, Regina Barzilay, and Tommi S. Jaakkola. Greed is good if randomized: New inference for dependency parsing. In EMNLP, 2014.","title":"Greed is good if randomized: New inference for dependency parsing","authors":"Yuan Zhang, Tao Lei, Regina Barzilay, Tommi S Jaakkola","year":"2014","venue":"EMNLP","doi":null,"url":null,"internal_url":null},{"position":36,"raw":"[Zhang et al., 2015] Yuan Zhang, Chengtao Li, Regina Barzi- lay, and Kareem Darwish. Randomized greedy inference for joint segmentation, POS tagging and dependency pars- ing. In NAACL-HLT, 2015.","title":"Randomized greedy inference for joint segmentation, POS tagging and dependency parsing","authors":"Yuan Zhang, Chengtao Li, Regina Barzilay, Kareem Darwish","year":"2015","venue":"NAACL-HLT","doi":null,"url":null,"internal_url":null}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}