[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-118777-en":3,"doc-seo-118777-105":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},118777,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697",8,"Research & Report","Neural networks as data - Doctor of Philosophy thesis","Neural networks are commonly used as models for classification, regression, or generative tasks, yet this thesis reframes them as data. Instead of representing signals as discrete arrays, continuous signals are modeled via implicit neural representations (INRs) that map coordinates to values, such as pixel locations to RGB. The work analyzes consequences and properties of this viewpoint, focusing on neural-network-based compression and generative modeling over datasets of INRs. It presents a unified neural codec using quantized INR weights for multiple modalities and studies learning distributions of INRs to support tasks like 3D shape inference and neural radiance field generation.","Neural networks as data  \nEmilien Dupont  \nJesus College University of Oxford  \nA thesis submitted for the degree of Doctor of Philosophy September 2022  \nAcknowledgements  \nFirst, I would like to sincerely thank my supervisors Yee Whye and Arnaud for their support and guidance throughout my DPhil. I am extremely grateful for the research freedom they gave me and for always encouraging me to explore any topic and idea I was interested in. I also want to thank the many amazing people I was lucky to collaborate with during my DPhil and internships, including the COIN group at Oxford, Qi, Josh, Miguel, Alex and Aditya at Apple and Hyunjik, Danilo, Ali and Dan at DeepMind.  \nI am also very grateful for all the friends that were there with me throughout the DPhil. There have been so many amazing people that I cannot possible list everyone.  \nTo everyone in the stats department and office 1 . 17 in particular. To Jef, Bobs, Jin, Kaspar, Robert, Faaiz, Edwin, Fran, Charline, Sheh, Hyunjik, Andrew, Eduard, Carlo, Desi, Tyler, Qinyi, Adam F, K and G, Emile and many more. Thankyou for making the office such a nice place to go to every day and for all the trips to the Japanese van and Gail’s.  \nI also want to thank the many sports groups that helped me relax and have fun during my DPhil. Thank you to the table tennis club for all the training sessions, matches and trips to Incha and Zhang Ji. To Eric, Kritica, Ali, Jef, Abe, Jinlin, Zhongyi, Seb, Virginia and many more. Thank you to the Wolfson tennis crew, I loved every session even when we were playing in zero degrees and snow. Thank you to the beach volleyball crew and the Sunday badminton group, for all the fun games and many dinners.  \nThank you also to many other friends in Oxford and outside of Oxford. In particular, Abdul, Kun, Lisa, Michael and William-thank you for being there forme. To Grace and Thu, thank you for making my internships so much more fun.  \nA special thank you to Jinlin for her continuous love and support-thank you for all the good times and for making the tough times a lot easier. Oxford life was better with you there. Finally, thank you to my family-Mor, Papa, Marie and Jesús-for always being there for me. Merci og tak.  \nAbstract  \nIn machine learning, neural networks are typically treated as models to be used for various tasks such as classification, regression or generative modeling. In this thesis, we take a different perspective and treat neural networks as data instead of models. Indeed, in deep learning, data is often represented by arrays, such as a 2D grid of pixels for images. However, the underlying signal represented by these arrays is often continuous, such as the scene depicted in an image. A powerful continuous alternative to discrete arrays is then to represent such signals with an implicit neural representation (INR), a neural network trained to output the appropriate signal value for any input spatial location. An image for example, can be parameterized by a neural network mapping pixel locations to RGB values. In this thesis, we investigate the consequences and compelling properties of using such neural networks as data. We motivate and explore this approach through two main applications: compression, where we store data as neural networks, and generative modeling, where we train generative models directly on datasets of neural networks. For data compression, we propose to store the quantized weights of an INRas a compressed code for a given signal (such as an image), instead of directly compressing the discretized signal. We show how this allows us to build a single neural codec that is seamlessly applicable to multiple data modalities, from images and audio to medical and climate data. For generative modeling, we propose to learn distributions of INRs, either implicitly by training on array data or explicitly  \nby directly training models on datasets of INRs. We demonstrate how such an approach leads to compelling algorithms for a range of","cbCaioHcS80nxplq","https://ap.wps.com/l/cbCaioHcS80nxplq","pdf",61192429,1,222,"English","en",105,"# Introduction\n## Motivation\n## Background\n## Properties of implicit neural representations\n## Thesis outline\n## Related concepts\n# Generative models as distributions of functions\n## Introduction\n## Representing data as functions\n## Learning distributions of functions","[{\"question\":\"What does it mean to treat neural networks as data instead of models?\",\"answer\":\"Neural networks are used to represent the underlying signal itself, rather than just mapping inputs to outputs as a predictive model. This perspective treats neural network parameters as the data representation.\"},{\"question\":\"How do implicit neural representations (INRs) relate to continuous signals?\",\"answer\":\"INRs use a neural network to output a signal value at any spatial coordinate, allowing continuous signals (e.g., an image) to be represented without relying on fixed discrete grids.\"},{\"question\":\"What two main applications does the thesis focus on?\",\"answer\":\"The thesis explores neural-network-based compression and generative modeling on datasets of neural networks. Compression is performed by storing quantized INR weights, while generative modeling learns distributions of INRs for tasks such as 3D shape inference and neural radiance field generation.\"}]","Neural networks as data - Doctor of Philosophy thesis | PDF",1785720196,559,{"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},"neural-networks-as-data-doctor-of-philosophy-thesis","",{"@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/neural-networks-as-data-doctor-of-philosophy-thesis/118777/",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},"What does it mean to treat neural networks as data instead of models?","Question",{"text":75,"@type":76},"Neural networks are used to represent the underlying signal itself, rather than just mapping inputs to outputs as a predictive model. This perspective treats neural network parameters as the data representation.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How do implicit neural representations (INRs) relate to continuous signals?",{"text":80,"@type":76},"INRs use a neural network to output a signal value at any spatial coordinate, allowing continuous signals (e.g., an image) to be represented without relying on fixed discrete grids.",{"name":82,"@type":73,"acceptedAnswer":83},"What two main applications does the thesis focus on?",{"text":84,"@type":76},"The thesis explores neural-network-based compression and generative modeling on datasets of neural networks. Compression is performed by storing quantized INR weights, while generative modeling learns distributions of INRs for tasks such as 3D shape inference and neural radiance field generation.","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"]