[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85389-en":3,"doc-seo-85389-105":29,"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":20,"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":13,"seo_description":14,"update_tm":27,"read_time":28},85389,1099513958607,"Jiven","https://ap-avatar.wpscdn.com/avatar/100002390cf8733938c?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778829742770036399",8,"Research & Report","Federated Topic Model and Model Pruning Based on Variational Autoencoder","Topic modeling supports discovering latent patterns and themes in large document collections, but cross-party analysis raises strict data-privacy requirements. Federated topic modeling enables multiple parties to jointly train while keeping local data protected, yet introduces communication overhead and performance loss. This paper proposes a federated variational-autoencoder topic model and accelerates it using neural model pruning driven by server-side pruning from client updates. Two pruning-rate strategies balance speed and accuracy, and experiments show faster training with preserved performance.","Federated Topic Model and Model Pruning Based on Variational  \nAutoencoder  \nChengjie Ma  \nBeijing Key Laboratory of Intelligent Communication Software and Multimedia, Beijing University of Posts and Telecommunications  \nBeijing, China  \nYawen Li∗ School of Economics and Management, Beijing University of Posts and Telecommunications Beijing, China  \nMeiyu Liang  \nBeijing Key Laboratory of Intelligent Communication Software and Multimedia, Beijing University of Posts and Telecommunications  \nBeijing, China  \nAng Li  \nBeijing Key Laboratory of Intelligent Communication Software and Multimedia, Beijing University of Posts and Telecommunications  \nBeijing, China  \narXiv :2311 .003 14v2 [ cs .LG] 11 Jul 2026  \nAbstract  \nTopic modeling has emerged as a valuable tool for discovering patterns and topics within large collections of documents. However, when cross-analysis involves multiple parties, data privacy becomes a critical concern. Federated topic modeling has been developed to address this issue, allowing multiple parties to jointly train models while protecting privacy. However, there are communication and performance challenges in the federated scenario. To solve these problems, this paper proposes a method to establish a federated topic model while ensuring the privacy of each node and uses neural network model pruning to accelerate the model. The client periodically sends cumulative neuron gradients and model weights to the server, and the server prunes the model. To address different requirements, two methods are proposed to determine the pruning rate. The first slowly prunes throughout training, which has limited acceleration during training but can ensure higher accuracy and significantly reduce inference time. The second quickly reaches the target pruning rate early in training and then continues training with a smaller model. This approach may lose more useful information but can complete training faster. Experimental results show that the proposed variational-autoencoder-based federated topic model pruning can greatly accelerate training while maintaining model performance.  \nKeywords  \nvariational autoencoder, topic model, federated learning, model pruning  \n1 Introduction  \nClassical topic models, including Latent Dirichlet Allocation (LDA) [1] and Probabilistic Latent Semantic Analysis (pLSA) [2], have been widely used to uncover latent themes in document collections. Subsequent work has extended topic-oriented analysis to heterogeneous semantic text mining [3], sentiment-spike explanation for evolving public events [4], and hashtag recommendation from multiple microblog features [5]. Dynamic-interest tracking further supports multi-view clustering of scholars whose research profiles change over time [6], while interpretable machine-learning  \n∗ [Corresponding author: warmly0716@126.com](Corresponding author: warmly0716@126.com).  \nmodels help expose the evidence behind intelligent decisions [7] . With the emergence of deep learning, neural topic models (NTMs) have gained popularity because they use neural networks to learn relationships between documents and topics, aiming for higher-quality topic representations.  \nIn fields such as science, technology, and innovation document analysis, topic models are used to compare topics in funded projects from different institutions and identify research strengths. Related representation techniques include heterogeneous graph attention for short-text classification [8] and deep modularity learning for community detection [9]. Scientific-publication representation can also combine semantic-similarity attention with hypergraph convolution [10] . For incomplete graphs, teacher–student distillation can recover information from missing features and structure [11], whereas retrieval-oriented masked autoencoding strengthens textual representations [12] . Self-supervised graph co-training provides another way to couple complementary relational signals [13], and filter-enhanced mult","cbCaioai7HY3R3cj","https://ap.wps.com/l/cbCaioai7HY3R3cj","pdf",653455,1,7,"English","en",105,"# Introduction\n## Classical Topic Models and Neural Topic Models\n## Federated Learning for Privacy-Preserving Modeling","[{\"question\":\"Why is federated learning used in topic modeling?\",\"answer\":\"Federated learning coordinates model training across multiple data owners without collecting their raw data, addressing privacy constraints in cross-party topic analysis.\"},{\"question\":\"How does the proposed federated variational-autoencoder approach handle communication and privacy?\",\"answer\":\"Clients periodically send cumulative neuron gradients and model weights to the server, while privacy is protected through the federated training protocol rather than sharing raw data.\"},{\"question\":\"What role does model pruning play in the proposed method?\",\"answer\":\"Model pruning is applied on the server side to accelerate the federated model, using two strategies to control pruning rate with trade-offs between training speed and information retention.\"}]",1784203075,18,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"federated-topic-model-and-model-pruning-based-on-variational-autoencoder","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/federated-topic-model-and-model-pruning-based-on-variational-autoencoder/85389/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is federated learning used in topic modeling?","Question",{"text":75,"@type":76},"Federated learning coordinates model training across multiple data owners without collecting their raw data, addressing privacy constraints in cross-party topic analysis.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed federated variational-autoencoder approach handle communication and privacy?",{"text":80,"@type":76},"Clients periodically send cumulative neuron gradients and model weights to the server, while privacy is protected through the federated training protocol rather than sharing raw data.",{"name":82,"@type":73,"acceptedAnswer":83},"What role does model pruning play in the proposed method?",{"text":84,"@type":76},"Model pruning is applied on the server side to accelerate the federated model, using two strategies to control pruning rate with trade-offs between training speed and information retention.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]