[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-125963-en":3,"doc-seo-125963-105":31,"detail-sidebar-cat-0-en-105":93},{"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":28,"seo_description":14,"update_tm":29,"read_time":30},125963,137451207643,"Noah","https://ap-avatar.wpscdn.com/davatar_3d24733baf745e90a7e4bdd5f77d97b2",8,"Research & Report","Towards Stable Machine Learning Model Retraining via Slowly Varying Sequences","The study addresses retraining machine learning (ML) models as new data batches become available, focusing on preserving stability of model structure rather than only maximizing accuracy per batch. It proposes a methodology that selects sequences of ML models stable across retraining iterations, by formulating a mixed-integer optimization guaranteed to recover Pareto-optimal solutions balancing predictive power and stability. Custom distance metrics encode interpretability-relevant consistency, supported by an efficient polynomial-time algorithm. Empirical evaluation shows improved stability with a controlled accuracy tradeoff, demonstrated in a real hospital case.","Towards Stable Machine Learning Model Retraining via Slowly Varying Sequences  \narXiv :2403 . 19871v4 [ cs .LG] 22 May 2024  \nDimitris Bertsimas  \nMIT  \nVassilis Digalakis Jr  \nHEC Paris  \nYu Ma  \nMIT  \nPhevos Paschalidis  \nHarvard University  \nAbstract  \nWe consider the task of retraining machine learning (ML) models when new batches of data become available. Existing methods focus largely on greedy approaches to find the best-performing model for each batch, without considering the stability of the model’s structure across retraining iterations. In this study, we propose a methodology for finding sequences of ML models that are stable across retraining iterations. We develop a mixed-integer optimization formulation that is guaranteed to recover Pareto optimal models (in terms of the predictive power-stability tradeoff) and an efficient polynomial-time algorithm that performs well in practice.  \nWe focus on retaining consistent analytical insights—which is important to model interpretability, ease of implementation, and fostering trust with users—by using custom-defined distance metrics that can be directly incorporated into the optimization problem. Our method shows stronger stability than greedily trained models with a small, controllable sacrifice in predictive power, as evidenced through a real-world case study in a major hospital system in Connecticut.  \n1 Introduction  \nThe adoption of ML models in high-stakes decision spaces such as healthcare [1] increases the need for interpretability and model understanding [2, 3, 4] . Expert users develop trust in AI-driven decision support systems based on both the systems’ alignment with the users’ personal expertise [5, 6] and the systems’ historical performance [7] . Importantly, trust is built over time and requires deep understanding of the underlying model and the resulting analytical insights [8] . In real-world production pipelines, ML models need to be frequently updated as more data becomes available [9] . Significant changes in the model’s structure and the resulting analytical insights [10] may lead to skepticism and hesitation for adoption. Especially in healthcare, such changes may challenge both physicians’ trust in AI systems and patients’ confidence in medical decision-making [11, 12], and may even lead to legal difficulties as a result of incidents of error [13] .  \nMultiple avenues have been proposed in the literature to address challenges associated with model retraining, including continual learning [14, 15], online learning [16], and others [17, 18, 19, 20] . Existing methods focus largely on maintaining the predictive power of the retrained models in the face of, e.g., data distributional shifts. Such methods do not, however, ensure that the decision path of a tree-based prediction or the important features selected by a model remain stable across different retraining iterations, and therefore do not guarantee retention of analytical insights. In this work, we explicitly incorporate this requirement into the (re)training process.  \nRetraining (batch learning) differs from online learning: the former periodically retrains the model from scratch with each new data batch; the latter continuously/incrementally updates the model as new data arrives. Thus, a batch learning-based retraining strategy allows for a globally optimal solution, as retraining uses the entire database, whereas online learning, often to ensure faster updates, incrementally incorporates only a portion of the data, hence providing greedier solutions [21] .  \nPreprint. Under review.  \nThe problem of developing stable ML models is not new in the statistics, computer science, and operations research literature [22, 10, 23] . Various methodologies have recently been proposed to either train structurally stable, interpretable models [24, 25, 26, 27], or black-box models that producestable feature interactions [28, 29] . In this paper, we develop a unifying, model-agnostic framework for training se","cbCaitd0ahDbbpMg","https://ap.wps.com/l/cbCaitd0ahDbbpMg","pdf",5155777,5,1,11,"English","en",105,"# Introduction\n## Model retraining vs online learning\n## Motivation for analytical insight stability\n# A Methodology for Retraining Machine Learning Models\n## Problem setting\n## Objectives and stability formulation","[{\"question\":\"What problem does the paper study in retraining machine learning models?\",\"answer\":\"It studies how to retrain ML models when new data batches arrive while maintaining stability of the model’s structure across retraining iterations, not just per-batch predictive performance.\"},{\"question\":\"How does the proposed method ensure stability across retraining steps?\",\"answer\":\"It introduces custom distance metrics and an optimization framework that explicitly balances predictive power and stability, recovering Pareto-optimal model sequences.\"},{\"question\":\"Does the method depend on a specific model type?\",\"answer\":\"No. The approach is model-agnostic and is demonstrated on logistic regression, classification trees, and boosted trees.\"}]","Towards Stable Machine Learning Model Retraining via Slowly Varying Sequences | PDF",1785902266,28,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":88,"head_meta":90,"extra_data":92,"updated_unix":29},"towards-stable-machine-learning-model-retraining-via-slowly-varying-sequences","",{"@graph":37,"@context":87},[38,55,70],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,49,52],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":48},"https://docshare.wps.com/document/","Document",2,{"item":50,"name":12,"@type":44,"position":51},"https://docshare.wps.com/document/research-report/",3,{"item":53,"name":13,"@type":44,"position":54},"https://docshare.wps.com/document/towards-stable-machine-learning-model-retraining-via-slowly-varying-sequences/125963/",4,{"url":53,"name":13,"@type":56,"author":57,"headline":13,"publisher":59,"fileFormat":62,"inLanguage":24,"description":14,"dateModified":63,"datePublished":64,"encodingFormat":62,"isAccessibleForFree":65,"interactionStatistic":66},"DigitalDocument",{"name":9,"@type":58},"Person",{"url":42,"name":60,"@type":61},"DocShare","Organization","application/pdf","2026-08-23","2026-08-05",true,{"@type":67,"interactionType":68,"userInteractionCount":20},"InteractionCounter",{"@type":69},"ViewAction",{"@type":71,"mainEntity":72},"FAQPage",[73,79,83],{"name":74,"@type":75,"acceptedAnswer":76},"What problem does the paper study in retraining machine learning models?","Question",{"text":77,"@type":78},"It studies how to retrain ML models when new data batches arrive while maintaining stability of the model’s structure across retraining iterations, not just per-batch predictive performance.","Answer",{"name":80,"@type":75,"acceptedAnswer":81},"How does the proposed method ensure stability across retraining steps?",{"text":82,"@type":78},"It introduces custom distance metrics and an optimization framework that explicitly balances predictive power and stability, recovering Pareto-optimal model sequences.",{"name":84,"@type":75,"acceptedAnswer":85},"Does the method depend on a specific model type?",{"text":86,"@type":78},"No. The approach is model-agnostic and is demonstrated on logistic regression, classification trees, and boosted trees.","https://schema.org",{"og:url":53,"og:type":89,"og:title":13,"og:site_name":60,"og:description":14},"article",{"robots":91,"canonical":53},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":94},[95,99,103,107,111,116,121,124,129,132,136],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":96,"show_sort_weight":97,"slug":98},"Story & Novel",90,"story-novel",{"id":48,"doc_module":4,"doc_module_name":47,"category_name":100,"show_sort_weight":101,"slug":102},"Literature",80,"literature",{"id":54,"doc_module":4,"doc_module_name":47,"category_name":104,"show_sort_weight":105,"slug":106},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":47,"category_name":108,"show_sort_weight":109,"slug":110},"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":47,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":47,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":47,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":47,"category_name":138,"show_sort_weight":20,"slug":139},19,"General","general"]