[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-121676-en":3,"doc-seo-121676-105":30,"detail-sidebar-cat-0-en-105":92},{"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":20,"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},121676,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697",8,"Research & Report","On The Impact of Machine Learning Randomness on Group Fairness","Group fairness metrics in machine learning quantify performance gaps across demographic groups, yet they often fluctuate sharply across training runs, undermining reliable empirical evaluation. The paper studies why this variance arises by examining multiple neural-network randomness sources during training. Results attribute fairness volatility to the learning process behavior of under-represented groups and identify stochasticity from data order as the dominant driver. A simple data-order modification for one epoch can control group-level accuracy with high efficiency and negligible overall performance impact.","On The Impact of Machine Learning Randomness on Group  \nFairness  \narXiv :2307 .04 138v 1 [ cs .LG] 9 Jul 2023  \nPrakhar Ganesh  \nNational University of Singapore Singapore [pganesh@comp.nus.edu.sg](pganesh@comp.nus.edu.sg)  \nMartin Strobel  \nNational University of Singapore Singapore [martin.r.strobel@gmail.com](martin.r.strobel@gmail.com)  \nHongyan Chang  \nNational University of Singapore Singapore [hongyan@comp.nus.edu.sg](hongyan@comp.nus.edu.sg)  \nReza Shokri  \nNational University of Singapore Singapore [reza@comp.nus.edu.sg](reza@comp.nus.edu.sg)  \nABSTRACT  \nStatistical measures for group fairness in machine learning reflect the gap in performance of algorithms across different groups. These measures, however, exhibit a high variance between different training instances, which makes them unreliable for empirical evaluation of fairness. What causes this high variance? We investigate the impact on group fairness of different sources of randomness in training neural networks. We show that the variance in group fairness measures is rooted in the high volatility of the learning process on under-represented groups. Further, we recognize the dominant source of randomness as the stochasticity of data order during training. Based on these findings, we show how one can control group-level accuracy (i.e., model fairness), with high efficiency and negligible impact on the model’s overall performance, by simply changing the data order for a single epoch.  \nCCS CONCEPTS  \n• Computing methodologies → Machine learning; • General and reference → Evaluation.  \nKEYWORDS  \nneural networks, fairness, randomness in training, evaluation  \nACM Reference Format:  \nPrakhar Ganesh, Hongyan Chang, Martin Strobel, and Reza Shokri. 2023. On The Impact of Machine Learning Randomness on Group Fairness. In 2023 ACM Conference on Fairness, Accountability, and Transparency (FAccT’23), June 12–15, 2023, Chicago, IL, USA. ACM, New York, NY, USA, 31 pages. [https://doi.org/10.1145/3593013.3594116](https://doi.org/10.1145/3593013.3594116)  \n1 INTRODUCTION  \nMachine learning models are shown to manifest and escalate historical biases present in their training data [1, 4, 16, 59]. Understanding these biases and the resulting ethical obligations have led to the rise offair machine learning research [13, 15, 37]. However, recent work  \nPermission to make digital or hard copies of part or all ofthis work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s) .  \nFAccT ’23, June 12–15, 2023, Chicago, IL, USA © 2023 Copyright held by the owner/author(s) .  \nACM ISBN 979-8-4007-0192-4/23/06 .  \n[https://doi.org/10.1145/3593013.3594116](https://doi.org/10.1145/3593013.3594116)  \nhas observed high variance in fairness measures across multiple training runs, usually attributed to non-determinism in training (e.g., weight initialization, data reshuffling, etc.) . These findings challenge the effectiveness of many bias mitigation algorithms [3, 48], and even the legitimacy of several fairness trends present in literature [51] . Thus, a reliable extraction of fairness measures requires accounting for the high variance due to randomness in the learning process to avoid lottery winners (see Fig. 1) .  \nThe standard solution to this concern is executing a large number of training runs with different randomness. However, such a solution creates huge computational demands when examining biases in neural networks. For instance, it costs about $450􀀠 to train a model of similar quality as GPT-3 [57], and thus executing multiple training runs of such a model is not practical. But, are multiple identical runs essential? Can we instead find an efficient alternative to measure this variance? Our paper answers this cr","cbCainF4Z0Qxz1sZ","https://ap.wps.com/l/cbCainF4Z0Qxz1sZ","pdf",2302417,1,31,"English","en",105,"# Abstract\n# Introduction\n## Motivation: fairness variance in training\n## Research questions\n## Key findings: dominant randomness and minority sensitivity\n## Approach: controlling group-level fairness via data order\n# Figures and experimental evidence","[{\"question\":\"What problem does the paper address regarding group fairness?\",\"answer\":\"It addresses the high variance of group fairness metrics across different training instances, which makes fairness evaluation unreliable without accounting for randomness in the learning process.\"},{\"question\":\"Which source of randomness is identified as dominant for fairness variance?\",\"answer\":\"The paper shows that stochasticity from data reshuffling (data order during training) dominates the observed variance in fairness measures.\"},{\"question\":\"How can group-level accuracy (model fairness) be controlled efficiently?\",\"answer\":\"By changing the data order for a single epoch, the method can control group-level performance with high efficiency and negligible impact on the model’s overall accuracy.\"}]","On The Impact of Machine Learning Randomness on Group Fairness | PDF",1785806155,78,{"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":87,"head_meta":89,"extra_data":91,"updated_unix":28},"on-the-impact-of-machine-learning-randomness-on-group-fairness","",{"@graph":36,"@context":86},[37,54,69],{"@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/on-the-impact-of-machine-learning-randomness-on-group-fairness/121676/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05","2026-08-04",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does the paper address regarding group fairness?","Question",{"text":76,"@type":77},"It addresses the high variance of group fairness metrics across different training instances, which makes fairness evaluation unreliable without accounting for randomness in the learning process.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"Which source of randomness is identified as dominant for fairness variance?",{"text":81,"@type":77},"The paper shows that stochasticity from data reshuffling (data order during training) dominates the observed variance in fairness measures.",{"name":83,"@type":74,"acceptedAnswer":84},"How can group-level accuracy (model fairness) be controlled efficiently?",{"text":85,"@type":77},"By changing the data order for a single epoch, the method can control group-level performance with high efficiency and negligible impact on the model’s overall accuracy.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]