[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117005-en":3,"doc-seo-117005-105":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},117005,16904993612988,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Efficient Machine Learning Software Stack from Algorithms to Compilation","Efficient Machine Learning Software Stack from Algorithms to Compilation analyzes how a complete machine learning software stack supports the full project life cycle, from problem definition and data processing to model design, frameworks, libraries, code optimization, and system management. It distinguishes algorithm design from compilation: algorithms target task performance, while compilation prioritizes execution time and resource usage on hardware, requiring arithmetic equivalence. The work addresses persistent efficiency gaps between desired performance and current solutions, highlighting opportunities across both design stages and their interaction.","Copyright by  \nZixuan Jiang 2023  \n1  \nThe Dissertation Committee for Zixuan Jiang certifies that this is the approved version of the following dissertation:  \nEfficient Machine Learning Software Stack from Algorithms to Compilation  \nCommittee:  \nDavid Zhigang Pan, Supervisor  \nDiana Marculescu  \n\n| Atlas Wang |\n| --- |\n| Qiang Liu |\n\nYuan Yu  \nEfficient Machine Learning Software Stack from Algorithms to Compilation  \nby  \nZixuan Jiang  \nDISSERTATION  \nPresented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment  \nof the Requirements  \nfor the Degree of  \nDOCTOR OF PHILOSOPHY  \nTHE UNIVERSITY OF TEXAS AT AUSTIN  \nAugust 2023  \nAcknowledgments  \nI sincerely thank my advisor, Professor David Z. Pan, for his guidance and support over the years. He is a wise leader who appropriately manages research projects and encourages me to pursue meaningful and impactful research problems. He is also a great mentor who guides my research, develops my skills for a future career, and cultivates my attitude toward healthy life. His warm advice and continuous support will positively impact my life in the future.  \nI would also like to thank other committee members. In particular, I want to thank Dr. Yuan Yu for his help during my internship in the ONNX Runtime team at Microsoft Cloud + AI. As a distinguished expert in machine learning systems, his insights help me understand the issues in the existing solutions. His energetic working style has long-term impacts on my future career and research. I also thank Professor Qiang Liu for his technical inspiration and suggestions on the mixed precision neural architecture search project. His enthusiasm for optimization techniques always encourages me to pursue more effective and elegant solutions. I would also like to express my thankfulness to Professor Diana Marculescu and Professor Atlas Wang for their kindness and support to this dissertation.  \nBesides, I really appreciate my colleagues during my internships. Dr. Yuan Yu, Sherlock Baihan Huang, and Dr. Wei Zuo were mentors for my internship at Microsoft Cloud + AI. They gave me a lot of precious advice on working with both industry and academia. Dr. Ebrahim M. Songhori, Shen Wang, Dr. Joe Wenjie Jiang, Dr. Azalia Mirhoseini, and  \nDr. Anna Goldie were mentors for my internship at Google Research, Brain Team. They offered precious insights and experiences in physical design algorithms and reinforcement learning. Dr. Lan Nie and Dr. Vineet Gupta were mentors for my internship at Google Ads. They taught me how to handle real-world problems with machine learning algorithms and infrastructures. Dr. Lifeng Nai, Dr. Safeen Huda, Dr. Sheng Li, and Dr. Jishen Zhao were mentors for my internship in the Google TPU team. They provided various domain-specific insights regarding machine learning systems and accelerators.  \nIn addition to my advisor, committee members, and colleagues, I am lucky to work and collaborate with many other people: Chengyue Gong, Dilin Wang at Computer Science Department of UT Austin, UTDA members and alums, including Dr. Yibo Lin, Dr. Meng Li, Dr. Biying Xu, Dr. Wuxi Li, Dr. Shounak Dhar, Dr. Zheng Zhao, Dr. Wei Ye, Dr. Mohamed Baker Alawieh, Dr. Wei Shi, Dr. Keren Zhu, Dr. Mingjie Liu, Dr. Jiaqi Gu, Dr. Chenghao Feng, Dr. Hao Chen, Ahmet F. Budak, Rachel S. Rajarathnam, Hanqing Zhu, Hyunsu Chae, Chen-Hao Hsu, et al. Their kindness in inspiring discussion and efforts for productive collaborations help develop and polish this dissertation.  \nLast, but not least, I am deeply thankful to my family. Without their support and sacrifice, it would have been impossible for me to finish this dissertation and pursue the PhD degree at UT Austin.  \nEfficient Machine Learning Software Stack from Algorithms to Compilation  \nPublication No.    \nZixuan Jiang, Ph.D.  \nThe University of Texas at Austin, 2023  \nSupervisor: David Zhigang Pan  \nMachine learning enables the extraction of knowledge from data a","cbCaig5VNuQ9n56O","https://ap.wps.com/l/cbCaig5VNuQ9n56O","pdf",3190720,1,204,"English","en",105,"# Acknowledgments\n# Dissertation Overview\n## Machine Learning Software Stack\n## Algorithm Design vs. Compilation\n## Ongoing Efficiency Challenges","[{\"question\":\"What does the machine learning software stack include in this dissertation?\",\"answer\":\"It covers components from problem definitions and data processing to model/method design, software frameworks and libraries, code optimization, and system management, supporting the full life cycle of a machine learning project.\"},{\"question\":\"How does algorithm design differ from compilation in the proposed framework?\",\"answer\":\"Algorithm design focuses on task-related performance, while compilation focuses on execution time and resource consumption on hardware. Compilation must preserve arithmetic equivalence to ensure consistent results.\"},{\"question\":\"Why is efficiency still a continuing challenge in current machine learning solutions?\",\"answer\":\"Despite innovations, the dissertation notes a gap between efficiency demands and existing solutions. It highlights that efficiency improvements are needed in both algorithm designs and compilation optimizations, as well as their interplay.\"}]","Efficient Machine Learning Software Stack from Algorithms to Compilation | PDF",1785673053,514,{"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},"efficient-machine-learning-software-stack-from-algorithms-to-compilation","",{"@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/efficient-machine-learning-software-stack-from-algorithms-to-compilation/117005/",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","2026-08-02",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},"What does the machine learning software stack include in this dissertation?","Question",{"text":75,"@type":76},"It covers components from problem definitions and data processing to model/method design, software frameworks and libraries, code optimization, and system management, supporting the full life cycle of a machine learning project.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does algorithm design differ from compilation in the proposed framework?",{"text":80,"@type":76},"Algorithm design focuses on task-related performance, while compilation focuses on execution time and resource consumption on hardware. Compilation must preserve arithmetic equivalence to ensure consistent results.",{"name":82,"@type":73,"acceptedAnswer":83},"Why is efficiency still a continuing challenge in current machine learning solutions?",{"text":84,"@type":76},"Despite innovations, the dissertation notes a gap between efficiency demands and existing solutions. It highlights that efficiency improvements are needed in both algorithm designs and compilation optimizations, as well as their interplay.","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"]