[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119033-en":3,"doc-seo-119033-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119033,137441390410,"Hazel","https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984",8,"Research & Report","Robustness of Algorithms for Causal Structure Learning to Hyperparameter Choice","Hyperparameters critically shape machine learning outcomes, yet causal structure learning makes their tuning especially difficult because it is unsupervised and lacks ground-truth graphs outside simulations. Prior research focused on performance comparisons of causal discovery methods, while the specific effects of hyperparameters on individual algorithms and on selecting the best algorithm for a given dataset remained insufficiently studied. This work empirically evaluates hyperparameter selection for seminal algorithms across datasets with varying complexity, showing that algorithm choice stays essential but hyperparameter selection in ensemble settings can strongly steer algorithm selection and enable or undermine state-of-the-art recovery.","Robustness of Algorithms for Causal Structure Learning to  \nHyperparameter Choice  \nDamian Machlanski D . MACHLANSKI @ESSEX . AC . UK  \nDepartment of Computer Science and Electronic Engineering, University of Essex  \nSpyridon Samothrakis SSAMOT @ESSEX . AC . UK  \nInstitute for Analytics and Data Science, University of Essex  \nPaul Clarke PCLARKE @ESSEX . AC . UK  \nInstitute for Social and Economic Research, University of Essex  \nEditors: Francesco Locatello and Vanessa Didelez  \nAbstract  \nHyperparameters play a critical role in machine learning. Hyperparameter tuning can make the difference between state-of-the-art and poor prediction performance for any algorithm, but it is particularly challenging for structure learning due to its unsupervised nature. As a result, hyperparameter tuning is often neglected in favour of using the default values provided by a particular implementation of an algorithm. While there have been numerous studies on performance evaluation of causal discovery algorithms, how hyperparameters affect individual algorithms, as well as the choice of the best algorithm for a specific problem, has not been studied in depth before.  \nThis work addresses this gap by investigating the influence of hyperparameters on causal structure learning tasks. Specifically, we perform an empirical evaluation of hyperparameter selection for some seminal learning algorithms on datasets of varying levels of complexity. We find that, while the choice of algorithm remains crucial to obtaining state-of-the-art performance, hyperparameter selection in ensemble settings strongly influences the choice of algorithm, in that a poor choice of hyperparameters can lead to analysts using algorithms which do not give state-of-the-art performance for their data.  \nKeywords: Hyperparameters, model selection, causal discovery, structure learning, performance evaluation, misspecification, robustness  \n1. Introduction  \nUncovering causal graphs is an immensely useful tool in data-driven decision-making as it helps understand the underlying data generating process. A large number of causal structure learning algorithms incorporate Machine Learning (ML) methods. These, in turn, heavily rely on hyperparameters (HPs) for accurate predictions (Bergstra et al., 2011) . In addition, there has been growing evidence that correctly specified HPs can close the performance gap between State-of-the-Art (SotA) and other methods (Paine et al., 2020 ; Zhang et al., 2021a ; Machlanski et al., 2023 ; Tnshoff et al., 2023) . Are hyperparameters as important in structure recovery?  \nHP optimisation is extremely challenging in structure learning as the true graphs are inaccessible outside of simulated environments. This inability to reliably tune could be one of the reasons behind the struggle to apply some of the algorithms to real data problems (Kaiser and Sipos, 2021), or why HPs are often completely neglected in this area. On the one hand, benchmarks and evaluation frameworks (e.g. Raghu et al. (2018); Tu et al. (2019)) usually focus on finding a learning algorithm that works best under specific circumstances but without considering HPs as part of the  \n© 2024 D. Machlanski, S. Samothrakis & P. Clarke.  \nMACHLANSKI SAMOTHRAKIS CLARKE  \nproblem. On the other hand, studies that address HP tuning (e.g. Strobl (2021); Biza et al. (2022)) consider individual algorithms but not the impact of tuning (or the lack of it) on selecting the best algorithm for the available data. Understanding how HPs affect algorithm choice, as well as individual methods, is clearly missing but can be a crucial next step towards more stable causal discovery in real data applications. To make matters worse, the evaluation metrics used for tuning can be imperfect and sometimes favour specific learning methods (Curth and van der Schaar, 2023) . This brings us to the core questions of this paper: Do different algorithms perform similarly given access to a hyperparameter oracle? How robust are ","cbCaihMenyV7HJQE","https://ap.wps.com/l/cbCaihMenyV7HJQE","pdf",2018632,1,37,"English","en",105,"# Introduction\n## Contributions\n## Related work\n# Abstract","[{\"question\":\"Why is hyperparameter tuning especially challenging in causal structure learning?\",\"answer\":\"Hyperparameter optimisation is hard because the true causal graphs are inaccessible in real settings and available only in simulated environments. This limits reliable tuning and often leads to neglecting hyperparameters in practice.\"},{\"question\":\"What does the study evaluate regarding hyperparameters and causal discovery algorithms?\",\"answer\":\"The work empirically investigates how hyperparameters influence graph recovery performance for individual causal structure learning algorithms and how hyperparameter selection affects the choice of the best algorithm.\"},{\"question\":\"What are the main findings about robustness to misspecified hyperparameters?\",\"answer\":\"The algorithm choice remains crucial for state-of-the-art performance, but hyperparameter selection in ensemble settings strongly influences algorithm choice. Poor hyperparameter choices can cause analysts to use algorithms that do not achieve state-of-the-art results for their data.\"}]","Robustness of Algorithms for Causal Structure Learning to Hyperparameter Choice | PDF",1785722014,93,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"robustness-of-algorithms-for-causal-structure-learning-to-hyperparameter-choice","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/robustness-of-algorithms-for-causal-structure-learning-to-hyperparameter-choice/119033/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is hyperparameter tuning especially challenging in causal structure learning?","Question",{"text":75,"@type":76},"Hyperparameter optimisation is hard because the true causal graphs are inaccessible in real settings and available only in simulated environments. This limits reliable tuning and often leads to neglecting hyperparameters in practice.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What does the study evaluate regarding hyperparameters and causal discovery algorithms?",{"text":80,"@type":76},"The work empirically investigates how hyperparameters influence graph recovery performance for individual causal structure learning algorithms and how hyperparameter selection affects the choice of the best algorithm.",{"name":82,"@type":73,"acceptedAnswer":83},"What are the main findings about robustness to misspecified hyperparameters?",{"text":84,"@type":76},"The algorithm choice remains crucial for state-of-the-art performance, but hyperparameter selection in ensemble settings strongly influences algorithm choice. Poor hyperparameter choices can cause analysts to use algorithms that do not achieve state-of-the-art results for their data.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]