[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119042-en":3,"doc-seo-119042-105":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},119042,137441390410,"Hazel","https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984",8,"Research & Report","Operating critical machine learning models in resource constrained regimes","Accelerated machine learning progress, especially deep learning, has enabled major advances in medical image analysis and computer-aided intervention, yet the training and deployment resource costs—data volume, compute, and energy—remain substantial barriers to clinic-wide adoption. The work examines resource efficiency approaches such as quantisation that reduce memory usage, while potentially affecting predictive performance. The focus is on the trade-off between resource consumption and performance for models used in critical clinical settings, supported by detailed experiments across resource metrics.","arXiv :2303 . 10 18 1v2 [ cs .LG] 4 Feb 2024  \nOperating critical machine learning models in resource constrained regimes  \nRaghavendra Selvan 1 ,2 , Julian Sch¨on 1 , and Erik B Dam 1  \n1 Department of Computer Science, University of Copenhagen  \n2 Department of Neuroscience, University of Copenhagen  \n[raghav@di.ku.dk](raghav@di.ku.dk)  \nAbstract. The accelerated development of machine learning methods, primarily deep learning, are causal to the recent breakthroughs in medical image analysis and computer aided intervention. The resource consumption of deep learning models in terms of amount of training data, compute and energy costs are known to be massive. These large resource costs can be barriers in deploying these models in clinics, globally. To address this, there are cogent efforts within the machine learning community to introduce notions of resource efficiency. For instance, using quantisation to alleviate memory consumption. While most of these methods are shown to reduce the resource utilisation, they could come at a cost in performance. In this work, we probe into the trade-off between resource consumption and performance, specifically, when dealing with models that are used in critical settings such as in clinics.3  \nKeywords: Resource efficiency · Image Classification · Deep Learning  \n1 Introduction  \nEvery third person on the planet does not yet have universal health coverage according to the estimates from the United Nations[30] . And improving access to universal health coverage is one of the UN Sustainable Development Goals. The use of machine learning (ML) based methods could be instrumental in achieving this objective. Reliable ML in clinical decision support could multiply the capabilities of healthcare professionals by reducing the time spent on several tedious and time-consuming steps. This can free up the precious time and effort of healthcare professionals to cater to the people in need[32] .  \nThe deployment of ML in critical settings like clinical environments is, however, currently in a nascent state. In addition to the fundamental challenges related to multi-site/multi-vendor validation, important issues pertaining to fairness, bias and ethics of these methods are still being investigated[24] . The demand for expensive material resources, in terms of computational infrastructure and energy requirements along with the climate impact of ML are also points of concern[25] .  \n3 Source Code: [https://github.com/raghavian/redl](https://github.com/raghavian/redl)  \n2 Selvan, Sch¨on and Dam  \nIn this work, we focus on the question of improving the resource efficiency of developing and deploying deep learning models for resource constrained environments [28,3] . Resource constraints in real world scenarios could be manifested asthe need for: low latency (corresponding to urgency), non-specialized hardware (instead of requiring GPU/similar infrastructure) and low-energy requirements (if models are deployed on portable settings) . The case for resource efficiency when deploying ML models is quite evident due to these aforementioned factors. In this work, we also argue that the training resource efficiency is important, as ML systems will be most useful in the clinical settings if they are able to continually learn [4,31] . We validate the usefulness of several methods for improving resource efficiency of deep learning models using detailed experiments. We show that resource efficiency – in terms of reducing compute time, GPU memory usage, energy consumption – can be achieved without degradation in performance for some classes of methods.  \n2 Methods for Resource Efficiency  \nIn this work we mainly consider reducing the overall memory footprint of deep learning models, which in turn could reduce computation time and the corresponding energy costs. We do not explicitly study the influence of model selection using efficient neural architecture search [29] or model compression using pruning or similar methods [","cbCaiqbdghNElJ2v","https://ap.wps.com/l/cbCaiqbdghNElJ2v","pdf",889113,1,12,"English","en",105,"# Introduction\n## Resource constraints in real-world clinical settings\n# Methods for Resource Efficiency\n## Reducing memory footprint and compute cost\n## Quantisation strategies","[{\"question\":\"Why are resource efficiency methods important for clinical deployment of deep learning models?\",\"answer\":\"Deep learning models require large training data, compute, and energy, which can block deployment in clinics. Resource-efficient approaches help meet constraints such as limited latency, non-specialized hardware, and low-energy operation.\"},{\"question\":\"What potential downside does quantisation introduce when improving resource efficiency?\",\"answer\":\"Quantisation reduces memory footprint and can lower compute cost, but it may come with loss in precision. In some cases, the added overhead of quantisation can offset performance or efficiency gains.\"},{\"question\":\"How does the work evaluate the relationship between resource consumption and performance?\",\"answer\":\"The study probes the trade-off between resource metrics—compute time, GPU memory usage, and energy consumption—and predictive performance for models used in critical clinical environments, using detailed experiments across quantisation-related methods.\"}]","Operating critical machine learning models in resource constrained regimes | PDF",1785722055,30,{"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},"operating-critical-machine-learning-models-in-resource-constrained-regimes","",{"@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/operating-critical-machine-learning-models-in-resource-constrained-regimes/119042/",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 are resource efficiency methods important for clinical deployment of deep learning models?","Question",{"text":75,"@type":76},"Deep learning models require large training data, compute, and energy, which can block deployment in clinics. Resource-efficient approaches help meet constraints such as limited latency, non-specialized hardware, and low-energy operation.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What potential downside does quantisation introduce when improving resource efficiency?",{"text":80,"@type":76},"Quantisation reduces memory footprint and can lower compute cost, but it may come with loss in precision. In some cases, the added overhead of quantisation can offset performance or efficiency gains.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the work evaluate the relationship between resource consumption and performance?",{"text":84,"@type":76},"The study probes the trade-off between resource metrics—compute time, GPU memory usage, and energy consumption—and predictive performance for models used in critical clinical environments, using detailed experiments across quantisation-related methods.","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},[93,97,101,105,110,115,120,122,127,130,134],{"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":29,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]