{"id":4212,"date":"2025-10-26T18:05:56","date_gmt":"2025-10-26T18:05:56","guid":{"rendered":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/?page_id=4212"},"modified":"2025-10-26T18:05:56","modified_gmt":"2025-10-26T18:05:56","slug":"dino-v2-for-self-supervised-rf-representations","status":"publish","type":"page","link":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/?page_id=4212","title":{"rendered":"DINO v2 for Self-Supervised RF Representations"},"content":{"rendered":"\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/wp-content\/uploads\/2025\/10\/DINO-v2-for-Self-Supervised-RF-Representations-Benjamin-J-Gilbert-Global-Midnight-Scan-Club.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of DINO v2 for Self-Supervised RF Representations Benjamin J Gilbert Global Midnight Scan Club.\"><\/object><a id=\"wp-block-file--media-599d930c-5e1e-4262-922c-ea2239187dda\" href=\"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/wp-content\/uploads\/2025\/10\/DINO-v2-for-Self-Supervised-RF-Representations-Benjamin-J-Gilbert-Global-Midnight-Scan-Club.pdf\">DINO v2 for Self-Supervised RF Representations Benjamin J Gilbert Global Midnight Scan Club<\/a><a href=\"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/wp-content\/uploads\/2025\/10\/DINO-v2-for-Self-Supervised-RF-Representations-Benjamin-J-Gilbert-Global-Midnight-Scan-Club.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-599d930c-5e1e-4262-922c-ea2239187dda\">Download<\/a><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">We adapt DINO-style self-supervised learning to<br>Wi-Fi channel state information (CSI) time-series data. By<br>treating the subcarrier\u2013time grid as a patchable signal and<br>training a Vision Transformer (ViT) with student\u2013teacher architecture, we learn RF embeddings that significantly improve<br>downstream decoding tasks over hand-crafted features. Our<br>method achieves superior linear-probe accuracy, produces wellclustered embedding geometries, and demonstrates strong data<br>efficiency across label fractions. We provide complete code and<br>reproducible build pipeline for RF self-supervised learning.<\/p>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/wp-content\/uploads\/2025\/10\/DINO-v2-for-Self-Supervised-RF-Representations-Benjamin-J-Gilbert-Global-Midnight-Scan-Club-Rev2.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of DINO v2 for Self-Supervised RF Representations Benjamin J Gilbert Global Midnight Scan Club Rev2.\"><\/object><a id=\"wp-block-file--media-c82c4b70-6959-464b-95b2-cd0a150ba44d\" href=\"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/wp-content\/uploads\/2025\/10\/DINO-v2-for-Self-Supervised-RF-Representations-Benjamin-J-Gilbert-Global-Midnight-Scan-Club-Rev2.pdf\">DINO v2 for Self-Supervised RF Representations Benjamin J Gilbert Global Midnight Scan Club Rev2<\/a><a href=\"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/wp-content\/uploads\/2025\/10\/DINO-v2-for-Self-Supervised-RF-Representations-Benjamin-J-Gilbert-Global-Midnight-Scan-Club-Rev2.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-c82c4b70-6959-464b-95b2-cd0a150ba44d\">Download<\/a><\/div>\n","protected":false},"excerpt":{"rendered":"<p>We adapt DINO-style self-supervised learning toWi-Fi channel state information (CSI) time-series data. Bytreating the subcarrier\u2013time grid as a patchable signal andtraining a Vision Transformer (ViT) with student\u2013teacher architecture, we learn RF embeddings that significantly improvedownstream decoding tasks over hand-crafted features. Ourmethod achieves superior linear-probe accuracy, produces wellclustered embedding geometries, and demonstrates strong dataefficiency across label&hellip;&nbsp;<\/p>\n","protected":false},"author":2,"featured_media":4076,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"neve_meta_sidebar":"","neve_meta_container":"","neve_meta_enable_content_width":"","neve_meta_content_width":0,"neve_meta_title_alignment":"","neve_meta_author_avatar":"","neve_post_elements_order":"","neve_meta_disable_header":"","neve_meta_disable_footer":"","neve_meta_disable_title":"","footnotes":""},"class_list":["post-4212","page","type-page","status-publish","has-post-thumbnail","hentry"],"_links":{"self":[{"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=\/wp\/v2\/pages\/4212","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4212"}],"version-history":[{"count":0,"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=\/wp\/v2\/pages\/4212\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=\/wp\/v2\/media\/4076"}],"wp:attachment":[{"href":"https:\/\/neurosphere-2.tail52f848.ts.net\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4212"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}