[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-183940-en":3,"doc-seo-183940-105":30,"detail-sidebar-cat-0-en-105":92},{"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":27,"seo_description":14,"update_tm":28,"read_time":29},183940,962084925636,"Sophia Brooks","https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d",8,"Research & Report","FederatedAveraging - Experimental Results","FederatedAveraging experiments compare federated optimization across image classification and language modeling tasks, focusing on how well FedAvg performs under practical mobile-device usability constraints. A proxy dataset supports an initial hyperparameter search while training more than 2000 models across configurations. Results are reported on CIFAR-10, and effectiveness is further tested on a real-world language modeling setup with natural client data partitioning. The study evaluates multiple model families including MNIST MLPs, CNNs, and an LSTM on IID versus pathological Non-IID client splits.","| Algorithm 1 FederatedAveraging. The K clients are indexed by k; B is the local minibatch size, E is the number of local epochs, and 􀀑 is the learning rate. |\n| --- |\n| Server executes:\u003Cbr>initialize w0\u003Cbr>for each round t = 1 ; 2 ; : : : dom max(C 􀀁 K; 1)\u003Cbr>St (random set of m clients) for each client k 2 St in parallel do\u003Cbr>wkt+1 ClientUpdate (k; wt) mt Pk2St nk\u003Cbr>wt+1 Pk2St ~~n~~m~~k~~t wkt+1 // Erratum4\u003Cbr>ClientUpdate(k; w): // Run on client k B (split Pk into batches of size B )\u003Cbr>for each local epoch i from 1 to E do for batch b 2 B do\u003Cbr>w w 􀀀 􀀑O` (w; b) return w to server |\n| 3 Experimental Results\u003Cbr>We are motivated by both image classification and language modeling tasks where good models can greatly enhance the usability of mobile devices. For each of these tasks we first picked a proxy dataset of modest enough size that we could thoroughly investigate the hyperparameters of the FedAvg algorithm. While each individual training run is relatively small, we trained over 2000 individual models for these experiments. We then present results on the benchmark CIFAR-10 image classification task. Finally, to demonstrate the effectiveness of FedAvg on a real-world problem with a natural partitioning of the data over clients, we evaluate on a large language modeling task.\u003Cbr>Our initial study includes three model families on two datasets. The first two are for the MNIST digit recognition task [26]: 1) A simple multilayer-perceptron with 2-hidden layers with 200 units each using ReLu activations (199,210 total parameters), which we refer to as the MNIST 2NN. 2) A CNN with two 5x5 convolution layers (the first with 32 channels, the second with 64, each followed with 2x2 max pooling), a fully connected layer with 512 units and ReLu activation, and a final softmax output layer (1,663,370 total parameters) . To study federated optimization, we also need to specify how the data is distributed over the clients. We study two ways of partitioning the MNIST data over clients: IID, where the data is shuffled, and then partitioned into 100 clients each receiving 600 examples, and Non-IID, where we first sort the data by digit label, divide it into 200 shards of size 300, and assign each of 100 clients 2 shards. This is a pathological non-IID partition of the data, as most clients |\n\n\n| 2NN\u003Cbr>C | ~~ ~~ IID ~~ ~~\u003Cbr>B = 1 B = 10 |  | ~~ ~~ NON-IID ~~ ~~\u003Cbr>B = 1 B = 10 |  |\n| --- | --- | --- | --- | --- |\n| 0.0 | 1455 | 316 | 4278 | 3275 |\n| 0.1 | 1474 (1:0􀀂) | 87 (3:6􀀂) | 1796 (2:4􀀂) | 664 (4:9􀀂) |\n| 0.2 | 1658 (0:9􀀂) | 77 (4:1􀀂) | 1528 (2:8􀀂) | 619 (5:3􀀂) |\n| 0.5 | — (—) | 75 (4:2􀀂) | — (—) | 443 (7:4􀀂) |\n| 1.0 | — (—) | 70 (4:5􀀂) | — (—) | 380 (8:6􀀂) |\n| CNN , E = 5 |  |  |  |  |\n| 0.0 | 387 | 50 | 1181 | 956 |\n| 0.1 | 339 (1:1􀀂) | 18 (2:8􀀂) | 1100 (1:1􀀂) | 206 (4:6􀀂) |\n| 0.2 | 337 (1:1􀀂) | 18 (2:8􀀂) | 978 (1:2􀀂) | 200 (4:8􀀂) |\n| 0.5 | 164 (2:4􀀂) | 18 (2:8􀀂) | 1067 (1:1􀀂) | 261 (3:7􀀂) |\n| 1.0 | 246 (1:6􀀂) | 16 (3:1􀀂) | — (—) | 97 (9:9􀀂) |\n\n\n| MNIST CNN , 99% ACCURACY |  |  |  |  |  |\n| --- | --- | --- | --- | --- | --- |\n| CNN | E B u |  |  | IID | NON-IID |\n| FED SGD | 1 1 1 |  |  | 626 | 483 |\n| FEDAV G | 5 1 5 |  |  | 179 (3:5􀀂) | 1000 (0:5􀀂) |\n| FEDAV G | 1 50 12 |  |  | 65 (9:6􀀂) | 600 (0:8􀀂) |\n| FEDAV G | 20 1 20 |  |  | 234 (2:7􀀂) | 672 (0:7􀀂) |\n| FEDAV G | 1 10 60 |  |  | 34 (18:4􀀂) | 350 (1:4􀀂) |\n| FEDAV G | 5 50 60 |  |  | 29 (21:6􀀂) | 334 (1:4􀀂) |\n| FEDAV G | 20 50 240 |  |  | 32 (19:6􀀂) | 426 (1:1􀀂) |\n| FEDAV G | 5 10 300 |  |  | 20 (31:3􀀂) | 229 (2:1􀀂) |\n| FEDAV G | 20 10 1200 |  |  | 18 (34:8􀀂) | 173 (2:8􀀂) |\n| LSTM | SHAKESPEARE LSTM ,\u003Cbr>E B u |  |  | 54% ACCURACY IID | NON-IID |\n| FED SGD\u003Cbr>FEDAV G\u003Cbr>FEDAV G\u003Cbr>FEDAV G\u003Cbr>FEDAV G\u003Cbr>FEDAV G | 1 1\u003Cbr>5 1\u003Cbr>5 5 | 1\u003Cbr>50\u003Cbr>1\u003Cbr>10\u003Cbr>50\u003Cbr>10 | 1.0 2488\u003Cbr>1.5 1635 (1:5􀀂)\u003Cbr>5.0 613 (4:1􀀂)\u003Cbr>7.4 460 (5:4􀀂)\u003Cbr>7.4 401 (6:2􀀂)\u003Cbr>37.1 192 (13:0􀀂) |  | 3906\u003Cbr>549 (7:1􀀂)\u003Cbr>597 (6:5􀀂)\u003Cbr>164 (23:8􀀂)\u003Cbr>152 (25:7􀀂)\u003Cbr>41 (95:3􀀂) |\n\n| ACC . | 80% | 82% | 85% |\n| --- | --- |","cbCaivwmsIq4rSSU","https://ap.wps.com/l/cbCaivwmsIq4rSSU","pdf",1311201,1,11,"English","en",105,"# Experimental Results\n## Experimental setup and datasets\n## Model families and client data partitioning\n## Reported benchmarks and accuracy trends","[{\"question\":\"What tasks are used to evaluate FedAvg in the experiments?\",\"answer\":\"The experiments target image classification and language modeling, including CIFAR-10 and a real-world language modeling task with natural client data partitioning.\"},{\"question\":\"How is data partitioning handled across clients?\",\"answer\":\"Two partitioning strategies are compared: IID partitions with shuffled data and Non-IID partitions that sort by label, shard the data, and assign shards to clients to create a pathological imbalance.\"},{\"question\":\"Which model families are included in the initial study?\",\"answer\":\"The initial study includes MNIST models: a two-hidden-layer multilayer perceptron (MNIST 2NN) and a two-convolution CNN, as well as a federated language modeling setup using an LSTM.\"}]","FederatedAveraging - Experimental Results | PDF",1788354568,28,{"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":87,"head_meta":89,"extra_data":91,"updated_unix":28},"federatedaveraging-experimental-results","",{"@graph":36,"@context":86},[37,54,69],{"@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/federatedaveraging-experimental-results/183940/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-09-06","2026-09-02",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What tasks are used to evaluate FedAvg in the experiments?","Question",{"text":76,"@type":77},"The experiments target image classification and language modeling, including CIFAR-10 and a real-world language modeling task with natural client data partitioning.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How is data partitioning handled across clients?",{"text":81,"@type":77},"Two partitioning strategies are compared: IID partitions with shuffled data and Non-IID partitions that sort by label, shard the data, and assign shards to clients to create a pathological imbalance.",{"name":83,"@type":74,"acceptedAnswer":84},"Which model families are included in the initial study?",{"text":85,"@type":77},"The initial study includes MNIST models: a two-hidden-layer multilayer perceptron (MNIST 2NN) and a two-convolution CNN, as well as a federated language modeling setup using an LSTM.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]