[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-290905-105":59,"doc-detail-290905-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","cmlcompiler-a-unified-compiler-for-classical-machine-learning-design-and-performance-study","CMLCompiler - A Unified Compiler for Classical Machine Learning - Design and Performance Study","","Classical machine learning (CML) forms a large portion of production machine-learning pipelines, yet it underutilizes modern state-of-the-practice devices and delivers limited portability and performance. Hybrid deployments combining deep learning (DL) and CML further face severe efficiency and deployment challenges. This paper introduces CMLCompiler for CML inference via unified operator representations and extended computational graphs, with graph conversion and optimization. Implemented on TVM, it shows up to 4.38× CPU, 3.31× GPU, and 5.09× IoT speedups versus scikit-learn and related tools, and up to 3.04× for mixed CML+DL pipelines.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/cmlcompiler-a-unified-compiler-for-classical-machine-learning-design-and-performance-study/290905/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/cmlcompiler-a-unified-compiler-for-classical-machine-learning-design-and-performance-study/290905.png","ImageObject",300,407,{"name":92,"@type":93},"Eden","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-01","2026-09-17",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",14,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does CMLCompiler address in classical machine learning deployment?","Question",{"text":113,"@type":114},"CML deployment suffers from portability and performance issues because existing CML frameworks rely on ad-hoc implementations for each model and device. Hybrid CML+DL deployments further worsen efficiency and portability problems.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does CMLCompiler unify CML inference to connect with DL compilers?",{"text":118,"@type":114},"It proposes two unified abstractions—operator representations and extended computational graphs—to convert and optimize CML into an output computational graph consumable by DL compilers or frameworks.",{"name":120,"@type":111,"acceptedAnswer":121},"What performance improvements does the evaluation report?",{"text":122,"@type":114},"Implemented on TVM, CMLCompiler achieves up to 4.38× speedup on CPU, 3.31× on GPU, and 5.09× on IoT devices compared with state-of-the-art solutions, and up to 3.04× speedup for mixed CML and DL pipelines.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},290905,1789642933,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":130,"read_time":41,"preview_page_count":4},1374391974468,"https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0","CMLCompiler: A Unified Compiler for Classical Machine  \nLearning  \narXiv :2301 . 13441v3 [ cs .LG] 28 Apr 2023  \nXu Wen  \nInstitute of Computing Technology, Chinese Academy of Sciences University of Chinese Academy of Sciences Beijing, China [wenxu@ict.ac.cn](wenxu@ict.ac.cn)  \nWanling Gao  \nInstitute of Computing Technology, Chinese Academy of Sciences University of Chinese Academy of Sciences Beijing, China [gaowanling@ict.ac.cn](gaowanling@ict.ac.cn)  \nAnzheng Li  \nInstitute of Computing Technology, Chinese Academy of Sciences University of Chinese Academy of Sciences Beijing, China [lianzheng20g@ict.ac.cn](lianzheng20g@ict.ac.cn)  \nLei Wang  \nInstitute of Computing Technology, Chinese Academy of Sciences University of Chinese Academy of Sciences Beijing, China [wanglei_2011@ict.ac.cn](wanglei_2011@ict.ac.cn)  \nZihan Jiang Huawei Technologies Co., Ltd.  \nBeijing, China [jiangzihan0512@gmail.com](jiangzihan0512@gmail.com)  \nJianfeng Zhan∗ Institute of Computing Technology, Chinese Academy of Sciences University of Chinese Academy of Sciences Beijing, China [zhanjianfeng@ict.ac.cn](zhanjianfeng@ict.ac.cn)  \nABSTRACT  \nClassical machine learning (CML) occupies nearly half of machine learning pipelines in production applications. Unfortunately, it fails to utilize the state-of-the-practice devices fully and performs poorly. Without a unified framework, the hybrid deployments of deep learning (DL) and CML also suffer from severe performance and portability issues. This paper presents the design of a unified compiler, called CMLCompiler, for CML inference. We propose two unified abstractions: operator representations and extended computational graphs. The CMLCompiler framework performs the conversion and graph optimization based on two unified abstractions, then outputs an optimized computational graph to DL compilers or frameworks. We implement CMLCompiler on TVM. The evaluation shows CMLCompiler’s portability and superior performance. It achieves up to 4.38× speedup on CPU, 3.31× speedupon GPU, and 5.09× speedup on IoT devices, compared to the stateof-the-art solutions — scikit-learn, intel sklearn, and hummingbird. Our performance of CML and DL mixed pipelines achievesup to 3.04x speedup compared with cross-framework implementations. The project documents and source code are available at [https://www.computercouncil.org/cmlcompiler](https://www.computercouncil.org/cmlcompiler).  \nCCS CONCEPTS  \n• Computing methodologies → Machine learning; • Computer systems organization → Real-time systems.  \nKEYWORDS  \nClassical Machine Learning, Deep Learning, Compiler  \n1 INTRODUCTION  \nDeep learning (DL) and classical machine learning (CML), collectively called machine learning (ML), have played an increasingly  \n∗ Corresponding author.  \nFigure 1: The CMLCompiler design. Our contributions are highlighted in green color.  \ncritical role in recent years. DL refers to those neural network models, such as convolutional neural networks (CNNs) [24], recurrent neural networks (RNNs) [28], and generative adversarial networks (GANs) [16] . Different from DL, CML represents a set of non-neural network models in ML, e.g., linear models [37], decision trees [26], random forests [4], and support vector machines [42] . DL stands out because of its accuracy, while CML is still widely used for lower time and energy costs. Doris Xin et al. [47] analyze 3000 production ML pipelines at Google and find that 40% of them use CML models. Besides, many real-world applications adopt hybrid deployments of CML and DL [2] to guarantee high accuracy and low latency [25, 27, 36, 38], e.g., DL models for feature embedding and CML models for classification or regression.  \nDL compilers, like TVM [7, 10, 23], provide a structural approach to tackle the portability issue and facilitates wide deployment of DL models on a broad spectrum of devices like GPUs, FPGAs, and IoT devices and guarantees an appreciable performance. DL compilers  \nXu Wen, Wanling Gao, Anzheng Li, Le","cbCaiuAcPB1UAw76","https://ap.wps.com/l/cbCaiuAcPB1UAw76","pdf",1265156,12,"English","# Abstract\n# 1 Introduction\n## Background: DL vs. CML and hybrid pipelines\n## Motivation: portability and performance gaps in CML frameworks\n## Approach: unifying CML abstractions with DL compiler infrastructure","[{\"question\":\"What problem does CMLCompiler address in classical machine learning deployment?\",\"answer\":\"CML deployment suffers from portability and performance issues because existing CML frameworks rely on ad-hoc implementations for each model and device. Hybrid CML+DL deployments further worsen efficiency and portability problems.\"},{\"question\":\"How does CMLCompiler unify CML inference to connect with DL compilers?\",\"answer\":\"It proposes two unified abstractions—operator representations and extended computational graphs—to convert and optimize CML into an output computational graph consumable by DL compilers or frameworks.\"},{\"question\":\"What performance improvements does the evaluation report?\",\"answer\":\"Implemented on TVM, CMLCompiler achieves up to 4.38× speedup on CPU, 3.31× on GPU, and 5.09× on IoT devices compared with state-of-the-art solutions, and up to 3.04× speedup for mixed CML and DL pipelines.\"}]","CMLCompiler - A Unified Compiler for Classical Machine Learning - Design and Performance Study | PDF"]