[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-203721-105":59,"doc-detail-203721-en":130},{"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":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","stitchnet-composing-neural-networks-from-pre-trained-fragments-abstract","StitchNet: Composing Neural Networks from Pre-Trained Fragments - Abstract","","StitchNet提出一种新的神经网络创建范式：从多个已预训练网络中抽取一段或多段连续层作为“片段”，再通过拼接构建目标任务所需网络。该方法利用Centered Kernel Alignment（CKA）度量片段兼容性，高效指导片段选择，并用反向传播训练完成组合网络。实验表明拼接后的网络在精度上可接近传统训练，同时显著降低计算与数据需求。文中进一步讨论支持个性化模型的按需生成与推理应用。",{"@graph":69,"@context":122},[70,84,105],{"@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":30,"@type":76,"position":81},"https://docshare.wps.com/document/technology/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/stitchnet-composing-neural-networks-from-pre-trained-fragments-abstract/203721/",{"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/stitchnet-composing-neural-networks-from-pre-trained-fragments-abstract/203721.png","ImageObject",300,407,{"name":92,"@type":93},"WPS_1786070896","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-03","2026-09-04",true,{"@type":102,"interactionType":103,"userInteractionCount":39},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"StitchNet的核心思想是什么？","Question",{"text":112,"@type":113},"StitchNet通过将多个预训练神经网络中的一段或多段连续层“片段”进行拼接来创建新网络，而不是从头训练或仅对现有模型进行重训。","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"StitchNet如何判断不同片段的兼容性？",{"text":117,"@type":113},"文中使用Centered Kernel Alignment（CKA）作为兼容性度量，用于高效指导片段选择并帮助组成高性能网络。",{"name":119,"@type":110,"acceptedAnswer":120},"StitchNet相较传统训练在资源需求上有什么优势？",{"text":121,"@type":113},"通过复用预训练片段并减少从头训练所需的计算与数据量，StitchNet能在精度接近传统方法的同时，用更少的计算与数据完成模型创建。","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},203721,1788562839,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":29,"category_name":30,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":39,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":39,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":129,"read_time":46},549768072016,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","StitchNet: Composing Neural Networks from  \nPre-Trained Fragments  \nSurat Teerapittayanon  \nNational Nanotechnology Center Pathum Thani, Thailand [surat.tee@nanotec.or.th](surat.tee@nanotec.or.th)  \nMarcus Comiter  \nHarvard University Cambridge, MA, USA [marcuscomiter@post.harvard.edu](marcuscomiter@post.harvard.edu)  \nBradley McDanel  \nFranklin & Marshall College Lancaster, PA, USA [bmcdanel@fandm.edu](bmcdanel@fandm.edu)  \nH.T. Kung  \nHarvard University Cambridge, MA, USA [kung@harvard.edu](kung@harvard.edu)  \nAbstract—We propose StitchNet, a novel neural network creation paradigm that stitches together fragments (one or more consecutive network layers) from multiple pre-trained neural networks. StitchNet allows the creation of high-performing neural networks without the large compute and data requirements needed under traditional model creation processes via backpropagation training. We leverage Centered Kernel Alignment (CKA) as a compatibility measure to efficiently guide the selection of these fragments in composing a network for a given task tailored to specific accuracy needs and computing resource constraints. We then show that these fragments can be stitched together to create neural networks with accuracy comparable to that of traditionally trained networks at a fraction of computing resource and data requirements. Finally, we explore a novel on-the-fly personalized model creation and inference application enabled by this new paradigm. The code is available at [https://github.com/steerapi/stitchnet](https://github.com/steerapi/stitchnet).  \nIndex Terms—StitchNet, Neural Networks, Deep Learning, Centered Kernel Alignment (CKA), Reusable Network Components  \nI. INTRODUCTION  \nAI models have become increasingly more complex to support additional functionality, multiple modalities, and higher accuracy. While the increased complexity has improved model utility and performance, it has imposed significant model training costs. Therefore, training complex models is often infeasible for resource-limited environments, such as those at the cloud edge and in fully or often disconnected environments. In response to these challenges, this paper proposes a new paradigm for creating neural networks: rather than training networks from scratch or retraining existing networks, we create neural networks through composition by stitching together fragments of existing pre-trained neural networks. A fragment is one or more consecutive layers of a neural network. We call the resulting neural network composed of one or more fragments a “StitchNet”(Fig. 1) . By significantly reducing the amount of computation and data resources needed to create neural networks, StitchNet enables an entire new set of applications, such as the rapid generation of personalized neural networks at the edge or in fully disconnected environments. StitchNet’s model creation mechanism is fundamentally different from today’s predominant backpropagation-based method for creating neural networks. Given a dataset anda task as input, the traditional training method uses backpropagation with stochastic gradient descent (SGD) or other  \nAlexNet ResNet DenseNet  \nExisting Networks Fragments StitchNets  \nFig. 1: Overview of the StitchNet approach. Existing networks (left) are cut into fragments (middle), which are composed into StitchNets (right) created for a particular task. No retraining is needed in this process.  \noptimization algorithms to adjust the weights of the network. This training process iterates through the full dataset multiple times, and therefore requires compute resources that scale with the amount of data and the complexity of the network. Training large models in this way also requires substantial amounts of data to mitigate overfitting. While successful, this traditional paradigm for model creation is not without its limitations, especially as AI moves out of the data center and into highly resource-constrained and disconnected environments. Creating com","cbCainFWA4vgZli7","https://ap.wps.com/l/cbCainFWA4vgZli7","pdf",608056,"English","# Abstract\n# Introduction","[{\"question\":\"StitchNet的核心思想是什么？\",\"answer\":\"StitchNet通过将多个预训练神经网络中的一段或多段连续层“片段”进行拼接来创建新网络，而不是从头训练或仅对现有模型进行重训。\"},{\"question\":\"StitchNet如何判断不同片段的兼容性？\",\"answer\":\"文中使用Centered Kernel Alignment（CKA）作为兼容性度量，用于高效指导片段选择并帮助组成高性能网络。\"},{\"question\":\"StitchNet相较传统训练在资源需求上有什么优势？\",\"answer\":\"通过复用预训练片段并减少从头训练所需的计算与数据量，StitchNet能在精度接近传统方法的同时，用更少的计算与数据完成模型创建。\"}]","StitchNet: Composing Neural Networks from Pre-Trained Fragments - Abstract | PDF"]