[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117944-en":3,"doc-seo-117944-105":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},117944,1649267921044,"Ava Thompson","https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1782875107921204101",8,"Research & Report","Neurosymbolic Approaches to Safe Machine Learning - Dissertation","Neurosymbolic Approaches to Safe Machine Learning investigates how to reason about neural-network-based systems that face substantial real-world safety risks. The research addresses limits of traditional program analysis, where neural networks are high-dimensional and existing verification techniques cannot scale, and testing challenges tied to adversarial examples that induce unsafe behavior. The work presents two complementary strategies: efficient neural-network robustness verification using machine learning-guided heuristics for program analysis, and shield-based designs that verify safety externally while preserving most neural performance benefits.","Copyright by  \nGreg Anderson 2023  \n1  \nThe Dissertation Committee for Greg Anderson certi􀀌es that this is the approved version of the following dissertation:  \nNeurosymbolic Approaches to Safe Machine Learning  \nCommittee:  \n\n| Swarat Chaudhuri, Supervisor |\n| --- |\n| I􀀘s􀀐l Dillig, Supervisor |\n| Joydeep Biswas |\n\nRajeev Alur  \nNeurosymbolic Approaches to Safe Machine Learning  \nby  \nGreg Anderson  \nDISSERTATION  \nPresented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Ful􀀌llment of the Requirements  \nfor the Degree of  \nDOCTOR OF PHILOSOPHY  \nTHE UNIVERSITY OF TEXAS AT AUSTIN  \nMay 2023  \nDedicated to my loving partner.  \n4  \nAcknowledgments  \nFirst, I want to thank my advisors, I􀀘s􀀐l Dillig and Swarat Chaudhuri, for all of their guidance and support over the years. I feel extremely lucky to have had two passionate, dedicated mentors to help me during my Ph.D. They have continually helped me shape and hone my research skills and keep my motivation throughout this process. Outside of research, both I􀀘s􀀐l and Swarat have cultivated a friendly, joyful experience, making a di􀀎cult journey more pleasant. I couldn't have asked for two better mentors to help me through my graduate studies.  \nMany times during my doctoral research I was honored to collaborate with excellent scientists and students. I want to recognize Xinyu Wang, Ken McMillan, Shankara Pailoor, Abhinav Verma, and Chenxi Yang for those wonderful collaborative opportunities. Each of them has improved my research skills and impacted my style through our interaction. I want to call out Xinyu in particular for helping me 􀀌nd my feet at the beginning of my Ph.D. and guiding me through my 􀀌rst experiences as a graduate student.  \nOutside of formal research collaborators, I have also bene􀀌tted immensely from all of my friends in the UToPiA and Trishul research groups. The community of these groups, both inside and outside the o􀀎ce was critical to my success and to my happiness during my time as a graduate student. I  \nwant to thank Kostas Ferles, Yuepeng Wang, Jia Chen, Yu Feng, Navid Yaghmazadeh, Jiayi Wei, Jon Stephens, Jocelyn Chen, Ben Mariano, Ben Sepanski, Ziteng Wang, Celeste Barnaby, Ian Kretz, Noah Patton, Anders Miltner, Calvin Smith, Dipak Chaudhari, Sam Anklesaria, Josh Ho􀀋man, Atharva Sehgal, Meghana Sistla, Yeming Wen, Dweep Trivedi, Eric Hsiung, Amitayush Thakur, Thomas Logan, and Christopher Hahn.  \nI am indebted to the members of my thesis committee, Joydeep Biswas and Rajeev Alur. Their feedback has been critical in shaping this dissertation and my work going forward into the future.  \nNone of this would have happened without the people who 􀀌rst introduced me to research. I want to express my appreciation for John Knight, my 􀀌rst research mentor, along with C􀀓esar Mu~noz and Aaron Dutle for introducing me to formal methods and program analysis. I credit these three, and my time at the University of Virginia and NASA, with turning me toward a career in research and encouraging me to pursue a Ph.D. in the 􀀌rst place.  \nFinally, I want to thank all of the people outside the research community who have kept me going for the past six years. Foremost among these is my partner, who provided boundless support over the course of my Ph.D. , and without whom this dissertation would never have been written. I would also like to thank my parents for setting me up for success in this endeavor and supporting me throughout. Lastly, the rest of my family and my friends in Austin deserve thanks for keeping me connected to the world outside of academia and for their support.  \nNeurosymbolic Approaches to Safe Machine Learning  \nPublication No.    \nGreg Anderson, Ph.D.  \nThe University of Texas at Austin, 2023  \nSupervisors: Swarat Chaudhuri  \nI􀀘s􀀐l Dillig  \nNeural networks have shown immense promise in solving a variety of challenging problems including computer vision, security, and robotic control. However these applications often co","cbCaijUubeh0vqqM","https://ap.wps.com/l/cbCaijUubeh0vqqM","pdf",5297048,1,168,"English","en",105,"# Acknowledgments\n# Abstract\n# Table of Contents","[{\"question\":\"What core safety problem does the dissertation address?\",\"answer\":\"Neural-network systems show strong real-world potential, but they introduce substantial risk that requires tools to analyze their behavior and ensure safe deployment.\"},{\"question\":\"How does the work first approach safe machine learning analysis?\",\"answer\":\"It develops an efficient method to verify neural-network robustness by using machine learning to create heuristics that improve the efficiency of existing program-analysis approaches.\"},{\"question\":\"What is the second major approach described in the dissertation?\",\"answer\":\"It uses a shield—an external, verifiable program—together with the neural network, combining them so the shield remains safe while retaining most of the neural network’s performance benefits.\"}]","Neurosymbolic Approaches to Safe Machine Learning - Dissertation | PDF",1785680474,423,{"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},"neurosymbolic-approaches-to-safe-machine-learning-dissertation","",{"@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/neurosymbolic-approaches-to-safe-machine-learning-dissertation/117944/",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 core safety problem does the dissertation address?","Question",{"text":75,"@type":76},"Neural-network systems show strong real-world potential, but they introduce substantial risk that requires tools to analyze their behavior and ensure safe deployment.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the work first approach safe machine learning analysis?",{"text":80,"@type":76},"It develops an efficient method to verify neural-network robustness by using machine learning to create heuristics that improve the efficiency of existing program-analysis approaches.",{"name":82,"@type":73,"acceptedAnswer":83},"What is the second major approach described in the dissertation?",{"text":84,"@type":76},"It uses a shield—an external, verifiable program—together with the neural network, combining them so the shield remains safe while retaining most of the neural network’s performance benefits.","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"]