{"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/copa-constrained-parafac2-for-sparse-large","title":"COPA: Constrained PARAFAC2 for Sparse & Large Datasets","arxiv_id":"1803.04572","date":"2018-03-12","proceeding":null,"authors":["Ardavan Afshar","Ioakeim Perros","Evangelos E. Papalexakis","Elizabeth Searles","Joyce Ho","Jimeng Sun"],"abstract":"PARAFAC2 has demonstrated success in modeling irregular tensors, where the\ntensor dimensions vary across one of the modes. An example scenario is modeling\ntreatments across a set of patients with the varying number of medical\nencounters over time. Despite recent improvements on unconstrained PARAFAC2,\nits model factors are usually dense and sensitive to noise which limits their\ninterpretability. As a result, the following open challenges remain: a) various\nmodeling constraints, such as temporal smoothness, sparsity and non-negativity,\nare needed to be imposed for interpretable temporal modeling and b) a scalable\napproach is required to support those constraints efficiently for large\ndatasets. To tackle these challenges, we propose a {\\it CO}nstrained {\\it\nPA}RAFAC2 (COPA) method, which carefully incorporates optimization constraints\nsuch as temporal smoothness, sparsity, and non-negativity in the resulting\nfactors. To efficiently support all those constraints, COPA adopts a hybrid\noptimization framework using alternating optimization and alternating direction\nmethod of multiplier (AO-ADMM). As evaluated on large electronic health record\n(EHR) datasets with hundreds of thousands of patients, COPA achieves\nsignificant speedups (up to 36 times faster) over prior PARAFAC2 approaches\nthat only attempt to handle a subset of the constraints that COPA enables.\nOverall, our method outperforms all the baselines attempting to handle a subset\nof the constraints in terms of speed, while achieving the same level of\naccuracy. Through a case study on temporal phenotyping of medically complex\nchildren, we demonstrate how the constraints imposed by COPA reveal concise\nphenotypes and meaningful temporal profiles of patients. The clinical\ninterpretation of both the phenotypes and the temporal profiles was confirmed\nby a medical expert.","url_abs":"http://arxiv.org/abs/1803.04572v2","url_pdf":"http://arxiv.org/pdf/1803.04572v2.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":"copa-constrained-parafac2-for-sparse-large","repo_url":"https://github.com/aafshar/COPA","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.04572","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}