[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-121777-en":3,"doc-seo-121777-105":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},121777,8796095360427,"Lucas Martin","https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d",6,"Technology","ProtoCaps - A Fast and Non-Iterative Capsule Network Routing Method","Capsule Networks deliver robust deep learning performance with fewer parameters than convolutional neural networks, yet their practical adoption is limited by slow iterative routing that repeatedly establishes connections between capsule layers and raises computational cost. The work presents ProtoCaps, a non-iterative routing mechanism inspired by trainable prototype clustering. A shared capsule subspace removes the need for projecting every lower-level capsule to every higher-level capsule, cutting memory usage during training. Experiments show improved results over existing non-iterative capsule methods on MNIST and Imagewoof, where iterative baselines struggle due to excessive GPU memory demand.","ProtoCaps: A Fast and Non-Iterative Capsule Network Routing Method  \nMiles Everett [m. everett.20@abdn. ac.uk](m. everett.20@abdn. ac.uk)  \nDepartment of Computing Science University of Aberdeen, UK  \nMingjun Zhong [mingjun.zhong@abdn. ac.uk](mingjun.zhong@abdn. ac.uk)  \nDepartment of Computing Science University of Aberdeen, UK  \nGeorgios Leontidis [georgios.leontidis@abdn. ac.uk](georgios.leontidis@abdn. ac.uk)  \nInterdisciplinary Centre for Data and AI Department of Computing Science University of Aberdeen, UK  \nReviewed on OpenReview: [https: // openreview. net/ forum? id= Id10mlBj cx](https: // openreview. net/ forum? id= Id10mlBj cx)  \nAbstract  \nCapsule Networks have emerged as a powerful class of deep learning architectures, known for robust performance with relatively few parameters compared to Convolutional Neural Networks (CNNs) . However, their inherent efficiency is often overshadowed by their slow, iterative routing mechanisms which establish connections between Capsule layers, posing computational challenges resulting in an inability to scale. In this paper, we introduce a novel, non-iterative routing mechanism, inspired by trainable prototype clustering. This innovative approach aims to mitigate computational complexity, while retaining, if not enhancing, performance efficacy. Furthermore, we harness a shared Capsule subspace, negating the need to project each lower-level Capsule to each higher-level Capsule, thereby significantly reducing memory requisites during training. Our approach demonstrates superior results compared to the current best non-iterative Capsule Network and tests on the Imagewoof dataset, which is too computationally demanding to handle efficiently by iterative approaches. Our findings underscore the potential of our proposed methodology in enhancing the operational efficiency and performance of Capsule Networks, paving the way for their application in increasingly complex computational scenarios. Code is available at [https://github.com/mileseverett/ProtoCaps](https://github.com/mileseverett/ProtoCaps).  \n1 Introduction  \nCapsule Networks (Sabour et al., 2017; Hinton et al., 2018; Ribeiro et al., 2020; Hahn et al., 2019; Ribeiro et al., 2022) are a family of Networks designed to overcome the shortcomings of CNNs being unable to handleequivariant translations. Capsule Networks do this by building parse trees of part-whole relationships using groups of individual neurons known as Capsules. While Capsule Networks have been successful in overcoming these shortcomings of CNNs, the core component of the majority of Capsule Networks is iterative, and thus slow and computationally inefficient to scale to images of a realistic size (example can be seen in Figure 1) .  \nMost work in Capsule Networks revolves around discovering new routing algorithms which are able to model the connections between Capsule from a lower layer i and a higher layer j, with these mainly being iterative in nature. These iterative algorithms come with the drawback that the already computationally and memoryintensive process of determining how much lower Capsules correspond to higher Capsules has to be repeated multiple times.  \nTest Accuracy  \n1.000  \n0.998  \n0.996  \n0.994  \n0.992  \n0.990  \nTest Accuracy vs FLOPs for Different Networks on MNIST  \n\n|  |  |  |  |  |  |  |  |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  |  |  | \u003Cbr>VB | \u003Cbr>Caps |  |  |\n| Prot | \u003Cbr>oCaps ( | ours) |  | SRCa | ps DRCaps |  |  |\n|  |  |  |  |  | EM | Caps |  |\n|  |  |  |  | \u003Cbr>ResNe | t |  |  |\n\n10m 15m 20m 25m 30m 35m 40m 45m 50m FLOPs (in Millions)  \n(a)  \nTest Accuracy  \n0.65  \n0.60  \n0.55  \n0.50  \n0.45  \n0.40  \n0.35  \n0.30  \n0.25  \n0.20  \nTest Accuracy vs FLOPs for Different Networks on Imagewoof  \n\n|  |  |  |\n| --- | --- | --- |\n| \u003Cbr>ProtoCa | ps (ours) |  |\n|  |  |  |\n|  |  |  |\n|  |  |  |\n|  |  |  |\n|  | ResNet\u003Cbr>\u003Cbr>SRCa |  |\n|  |  | ps |\n|  |  |  |\n\n100m 1000m  \nFLOPs (in Millions)  \n(b)  \nFigure 1: Comparative analysis o","cbCair6si7xPGp07","https://ap.wps.com/l/cbCair6si7xPGp07","pdf",1071513,1,13,"English","en",105,"# Introduction\n## Background on Capsule Networks\n## Routing algorithms and computational bottlenecks\n## Proposed ProtoCaps approach","[{\"question\":\"Why are iterative routing mechanisms a problem for Capsule Networks?\",\"answer\":\"Iterative routing repeatedly computes capsule-to-capsule connections across layers, which increases runtime and memory usage, making scaling to realistic image sizes difficult.\"},{\"question\":\"What is ProtoCaps and how does it reduce routing cost?\",\"answer\":\"ProtoCaps introduces a non-iterative routing mechanism based on trainable prototype clustering, which mitigates computational complexity and improves efficiency.\"},{\"question\":\"How does ProtoCaps reduce training memory requirements?\",\"answer\":\"It uses a shared capsule subspace, avoiding the need to project each lower-level capsule representation to every higher-level capsule representation.\"}]","ProtoCaps - A Fast and Non-Iterative Capsule Network Routing Method | PDF",1785806793,33,{"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},"protocaps-a-fast-and-non-iterative-capsule-network-routing-method","",{"@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/technology/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/protocaps-a-fast-and-non-iterative-capsule-network-routing-method/121777/",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","2026-08-04",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},"Why are iterative routing mechanisms a problem for Capsule Networks?","Question",{"text":75,"@type":76},"Iterative routing repeatedly computes capsule-to-capsule connections across layers, which increases runtime and memory usage, making scaling to realistic image sizes difficult.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is ProtoCaps and how does it reduce routing cost?",{"text":80,"@type":76},"ProtoCaps introduces a non-iterative routing mechanism based on trainable prototype clustering, which mitigates computational complexity and improves efficiency.",{"name":82,"@type":73,"acceptedAnswer":83},"How does ProtoCaps reduce training memory requirements?",{"text":84,"@type":76},"It uses a shared capsule subspace, avoiding the need to project each lower-level capsule representation to every higher-level capsule representation.","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},[93,97,101,105,110,113,118,123,128,131,135],{"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":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":111,"slug":112},50,"technology",{"id":114,"doc_module":4,"doc_module_name":46,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":119,"doc_module":4,"doc_module_name":46,"category_name":120,"show_sort_weight":121,"slug":122},8,"Research & Report",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"]