{"id":33865,"date":"2019-10-31T21:55:06","date_gmt":"2019-10-31T18:55:06","guid":{"rendered":"https:\/\/prohoster.info\/blog\/microsoft-otkryl-kod-biblioteki-vektornogo-poiska-ispolzuemoj-v-bing\/"},"modified":"2019-10-31T21:55:06","modified_gmt":"2019-10-31T18:55:06","slug":"microsoft-otkryl-kod-biblioteki-vektornogo-poiska-ispolzuemoj-v-bing","status":"publish","type":"post","link":"https:\/\/prohoster.info\/da\/blog\/news\/microsoft-otkryl-kod-biblioteki-vektornogo-poiska-ispolzuemoj-v-bing","title":{"rendered":"Microsoft har \u00e5bnet kildekoden til den vektors\u00f8gningsbibliotek, der bruges i Bing.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Microsoft offentliggjorde<\/a><\/noindex> originale tekster fra maskinl\u00e6ringsbiblioteket SPTAG<\/a><\/noindex> (Space Partition Tree And Graph) med implementering af algoritmen til tiln\u00e6rmet n\u00e6rmeste nabo s\u00f8gning<\/a><\/noindex>. Biblioteket er udviklet<\/a><\/noindex> i Microsoft Researchs forskningsenhed og i teknologiudviklingscentret for s\u00f8gning (Microsoft Search Technology Center). I praksis anvendes SPTAG i s\u00f8gemaskinen Bing til at bestemme de mest relevante resultater under hensyntagen til konteksten af s\u00f8geforesp\u00f8rgsler. Koden er skrevet i C++ og distribueres<\/a><\/noindex> under MIT-licensen. Der tilbydes bygning for Linux og Windows. Der er en binding til Python.<\/p>\n

Selvom ideerne om at anvende vektorlager i s\u00f8gemaskiner har eksisteret i et stykke tid, er implementeringen i praksis blevet hindret af den store ressourcekrav til operationer med vektorer og begr\u00e6nsninger i skalerbarhed. Kombinationen af metoder fra dyb l\u00e6ring med algoritmer til tiln\u00e6rmet n\u00e6rmeste nabo s\u00f8gning har gjort det muligt at bringe ydeevnen og skalerbarheden i vektorsystemer op p\u00e5 et niveau, der er acceptabelt for store s\u00f8gemaskiner. For eksempel, i Bing, for et vektorreferencelager p\u00e5 over 150 milliarder vektorer, ligger tiden for at hente de mest relevante resultater inden for 8 ms.<\/p>\n

Biblioteket inkluderer v\u00e6rkt\u00f8jer til opbygning af indekser og organisering af s\u00f8gning i vektorer samt et s\u00e6t v\u00e6rkt\u00f8jer til vedligeholdelse af et distribueret online-s\u00f8gesystem, der d\u00e6kker meget store samlinger af vektorer. Der tilbydes<\/a><\/noindex> f\u00f8lgende moduler: index builder til indeksering, searcher til s\u00f8gning ved hj\u00e6lp af indekset, distribueret i et klynge af flere noder, server til at k\u00f8re h\u00e5ndterere p\u00e5 noderne, Aggregator til at kombinere flere servere til et hele og klient til at sende foresp\u00f8rgsler. Der underst\u00f8ttes tilf\u00f8jelse af nye vektorer til indekset og fjernelse af vektorer on-the-fly.<\/p>\n

Biblioteket foruds\u00e6tter, at de behandlede og pr\u00e6senterede data i samlingen er opstillet som relaterede vektorer, der kan sammenlignes p\u00e5 basis af euclidian<\/a><\/noindex> (L2) eller kosinus<\/a><\/noindex> afstande. Ved en s\u00f8geforesp\u00f8rgsel returneres vektorer, hvis afstand til den originale vektor er minimal. I SPTAG tilbydes der to metoder til organisering af vektorplads: SPTAG-KDT (K-dimensionelt tr\u00e6 (kd-tree<\/a><\/noindex>) og graf over relative naboer<\/a><\/noindex>) og SPTAG-BKT (k-centroid tr\u00e6 (k-means tree<\/a><\/noindex> og og grafik over relative n\u00e6rheder). Den f\u00f8rste metode kr\u00e6ver f\u00e6rre ressourcer, n\u00e5r den arbejder med indekset, mens den anden viser en h\u00f8jere pr\u00e6cision i s\u00f8gningerne ved meget store samlinger af vektorer.<\/p>\n

Vektor-s\u00f8gning er ikke begr\u00e6nset til tekst og kan ogs\u00e5 bruges til multimedieinformation og billeder samt i systemer til automatisk generering af anbefalinger. For eksempel blev der i en prototype baseret p\u00e5 PyTorch-rammen implementeret et vektorsystem til s\u00f8gning baseret p\u00e5 lighed mellem objekter i billeder, konstrueret ved hj\u00e6lp af data fra flere referencekollektioner med billeder af dyr, katte og hunde, der blev konverteret til vektors\u00e6t. N\u00e5r et indkommende billede skal s\u00f8ges, omdannes det ved hj\u00e6lp af en maskinl\u00e6ringsmodel til en vektor, baseret p\u00e5 hvilken algoritmen SPTAG v\u00e6lger de mest lignende vektorer fra indekset, og som resultat returneres de associerede billeder.<\/p>\n

<\/a><\/noindex><\/p>\n

Kilde: opennet.ru<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u041a\u043e\u043c\u043f\u0430\u043d\u0438\u044f Microsoft \u043e\u043f\u0443\u0431\u043b\u0438\u043a\u043e\u0432\u0430\u043b\u0430 \u0438\u0441\u0445\u043e\u0434\u043d\u044b\u0435 \u0442\u0435\u043a\u0441\u0442\u044b \u0431\u0438\u0431\u043b\u0438\u043e\u0442\u0435\u043a\u0438 \u043c\u0430\u0448\u0438\u043d\u043d\u043e\u0433\u043e \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u044f SPTAG (Space Partition Tree And Graph) \u0441 \u0440\u0435\u0430\u043b\u0438\u0437\u0430\u0446\u0438\u0435\u0439 \u0430\u043b\u0433\u043e\u0440\u0438\u0442\u043c\u0430 \u043f\u0440\u0438\u0431\u043b\u0438\u0437\u0438\u0442\u0435\u043b\u044c\u043d\u043e\u0433\u043e \u043f\u043e\u0438\u0441\u043a\u0430 \u0431\u043b\u0438\u0436\u0430\u0439\u0448\u0435\u0433\u043e \u0441\u043e\u0441\u0435\u0434\u0430. \u0411\u0438\u0431\u043b\u0438\u043e\u0442\u0435\u043a\u0430 \u0440\u0430\u0437\u0440\u0430\u0431\u043e\u0442\u0430\u043d\u0430 \u0432 \u0438\u0441\u0441\u043b\u0435\u0434\u043e\u0432\u0430\u0442\u0435\u043b\u044c\u0441\u043a\u043e\u043c \u043f\u043e\u0434\u0440\u0430\u0437\u0434\u0435\u043b\u0435\u043d\u0438\u0438 Microsoft Research \u0438 \u0446\u0435\u043d\u0442\u0440\u0435 \u0440\u0430\u0437\u0440\u0430\u0431\u043e\u0442\u043a\u0438 \u043f\u043e\u0438\u0441\u043a\u043e\u0432\u044b\u0445 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0439 (Microsoft Search Technology Center). \u041d\u0430 \u043f\u0440\u0430\u043a\u0442\u0438\u043a\u0435 SPTAG \u043f\u0440\u0438\u043c\u0435\u043d\u044f\u0435\u0442\u0441\u044f \u0432 \u043f\u043e\u0438\u0441\u043a\u043e\u0432\u043e\u0439 \u0441\u0438\u0441\u0442\u0435\u043c\u0435 Bing \u0434\u043b\u044f \u043e\u043f\u0440\u0435\u0434\u0435\u043b\u0435\u043d\u0438\u044f \u043d\u0430\u0438\u0431\u043e\u043b\u0435\u0435 \u0440\u0435\u043b\u0435\u0432\u0430\u043d\u0442\u043d\u044b\u0445 \u0440\u0435\u0437\u0443\u043b\u044c\u0442\u0430\u0442\u043e\u0432 \u0441 \u0443\u0447\u0451\u0442\u043e\u043c \u043a\u043e\u043d\u0442\u0435\u043a\u0441\u0442\u0430 \u043f\u043e\u0438\u0441\u043a\u043e\u0432\u044b\u0445 \u0437\u0430\u043f\u0440\u043e\u0441\u043e\u0432. […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-33865","post","type-post","status-publish","format-standard","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47Microsoft har \u00e5bnet koden til vektor-s\u00f8gebiblioteket, der bruges i Bing | ProHoster","description":"Microsoft har offentliggjort kildekoden til maskinl\u00e6ringsbiblioteket","canonical_url":"https:\/\/prohoster.info\/da\/blog\/news\/microsoft-otkryl-kod-biblioteki-vektornogo-poiska-ispolzuemoj-v-bing","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"da_DK","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47Microsoft \u043e\u0442\u043a\u0440\u044b\u043b \u043a\u043e\u0434 \u0431\u0438\u0431\u043b\u0438\u043e\u0442\u0435\u043a\u0438 \u0432\u0435\u043a\u0442\u043e\u0440\u043d\u043e\u0433\u043e \u043f\u043e\u0438\u0441\u043a\u0430, \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u0443\u0435\u043c\u043e\u0439 \u0432 Bing | ProHoster","og:description":"\u041a\u043e\u043c\u043f\u0430\u043d\u0438\u044f Microsoft \u043e\u043f\u0443\u0431\u043b\u0438\u043a\u043e\u0432\u0430\u043b\u0430 \u0438\u0441\u0445\u043e\u0434\u043d\u044b\u0435 \u0442\u0435\u043a\u0441\u0442\u044b \u0431\u0438\u0431\u043b\u0438\u043e\u0442\u0435\u043a\u0438 \u043c\u0430\u0448\u0438\u043d\u043d\u043e\u0433\u043e \u043e\u0431\u0443\u0447\u0435\u043d\u0438\u044f","og:url":"https:\/\/prohoster.info\/da\/blog\/news\/microsoft-otkryl-kod-biblioteki-vektornogo-poiska-ispolzuemoj-v-bing","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2019-10-31T18:55:06+00:00","article:modified_time":"2019-10-31T18:55:06+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"33865","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":"2026-01-21 17:01:19","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-03-01 02:32:26","updated":"2026-01-21 17:01:19","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/posts\/33865","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/comments?post=33865"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/posts\/33865\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/media?parent=33865"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/categories?post=33865"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/da\/wp-json\/wp\/v2\/tags?post=33865"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}