{"id":54780,"date":"2020-01-04T00:00:00","date_gmt":"2020-01-03T21:00:00","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/nejroseti-kuda-eto-vse-dvizhetsya"},"modified":"2020-02-18T14:02:50","modified_gmt":"2020-02-18T11:02:50","slug":"nejroseti-kuda-eto-vse-dvizhetsya","status":"publish","type":"post","link":"https:\/\/prohoster.info\/lv\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","title":{"rendered":"Neironu t\u012bkli. Kur viss tas virz\u0101s","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>Raksts sast\u0101v no div\u0101m da\u013c\u0101m:<\/p>\n<p><\/p>\n<ol>\n<li>\u012ass da\u017eu arhitekt\u016bru apraksts t\u012bklos, kas atkl\u0101j objektus att\u0113l\u0101 un att\u0113lu segment\u0101cij\u0101 ar visiem saprotam\u0101kajiem resursu sai\u0161u piem\u0113riem. Centos izv\u0113l\u0113ties video skaidrojumus, v\u0113lams krievu valod\u0101.<\/li>\n<li>Otr\u0101 da\u013ca ir centiens apzin\u0101t neironu t\u012bklu arhitekt\u016bru att\u012bst\u012bbas virzienus. Un tehnolo\u0123ijas, kas balst\u012btas uz t\u0101m.<\/li>\n<\/ol>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neironu t\u012bkli. Kur viss tas virz\u0101s\" src=\"\/wp-content\/uploads\/2020\/01\/3e0238547dc956dbb069c11241e4534f.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Att\u0113ls 1 \u2013 Izprast neironu t\u012bklu arhitekt\u016bras nav viegli<\/p>\n<p><\/p>\n<p>Viss s\u0101k\u0101s ar to, ka izveidoju divas demonstr\u0101cijas lietotnes att\u0113lu klasifik\u0101cijai un objektu atkl\u0101\u0161anai Android telefon\u0101:<\/p>\n<p><\/p>\n<ul>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/junkyard\/tree\/master\/object_classifier\">Back-end demonstr\u0101cija<\/a><\/noindex>, kad dati tiek apstr\u0101d\u0101ti server\u012b un p\u0101rs\u016bt\u012bti uz t\u0101lruni. Att\u0113lu klasifik\u0101cija (image classification) tr\u012bs l\u0101\u010du veidiem: br\u016bnais, melnais un m\u012bkstais.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/android\/tree\/master\/object_detection_demo\">Front-end demonstr\u0101cija<\/a><\/noindex>, kad dati tiek apstr\u0101d\u0101ti pa\u0161\u0101 telefon\u0101. Objektu atkl\u0101\u0161ana (object detection) tr\u012bs veidiem: lazdu rieksts, v\u012b\u0123es un dateles.<\/li>\n<\/ul>\n<p><noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex><\/p>\n<p>Ir at\u0161\u0137ir\u012bba starp att\u0113lu klasifik\u0101cijas uzdevumiem, objektu atkl\u0101\u0161anu att\u0113l\u0101 un <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/analytics-vidhya\/image-classification-vs-object-detection-vs-image-segmentation-f36db85fe81\">att\u0113lu segment\u0101ciju<\/a><\/noindex>. T\u0101p\u0113c rad\u0101s nepiecie\u0161am\u012bba uzzin\u0101t, kuras neironu t\u012bklu arhitekt\u016bras atkl\u0101j objektus att\u0113los un kuras var veikt segment\u0101ciju. Atradu \u0161\u0101das arhitekt\u016bras piem\u0113rus ar visiem saprotam\u0101kajiem resursu sai\u0161u piem\u0113riem:<\/p>\n<p><\/p>\n<ul>\n<li>R-CNN arhitekt\u016bru s\u0113rija (<strong>R<\/strong>egions with <strong>C<\/strong>onvolution <strong>N<\/strong>eural <strong>N<\/strong>etworks features): R-CNN, Fast R-CNN, <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/@smallfishbigsea\/faster-r-cnn-explained-864d4fb7e3f8\">Faster R-CNN<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/0vt05rQqk_I\">Mask R-CNN<\/a><\/noindex>. Lai atkl\u0101tu objektu att\u0113l\u0101, izmantojot Region Proposal Network (RPN) meh\u0101nismu, izdal\u0101s ierobe\u017eoti re\u0123ioni (bounding boxes). S\u0101kotn\u0113ji RPN viet\u0101 tika izmantots l\u0113n\u0101ks meh\u0101nisms Selective Search. P\u0113c tam izdal\u012btie ierobe\u017eotie re\u0123ioni tiek iesniegti parastai neironu t\u012bklam klasifik\u0101cijai. R-CNN arhitekt\u016br\u0101 ir skaidri \u201cfor\u201d cikli ierobe\u017eoto re\u0123ionu p\u0101rl\u016bko\u0161anai, kopum\u0101 l\u012bdz 2000 iter\u0101cij\u0101m caur iek\u0161\u0113jo AlexNet t\u012bklu. Skaitlis skaidri \u201cfor\u201d ciklu pal\u0113nina att\u0113lu apstr\u0101des \u0101trumu. Skaitlis skaidri \u201cfor\u201d ciklu, iter\u0101ciju pa iek\u0161\u0113jo neironu t\u012bklu, samazin\u0101s ar katru jauno arhitekt\u016bras versiju, un tiek veikti desmitiem citu izmai\u0146u \u0101truma palielin\u0101\u0161anai un uzdevuma atkl\u0101\u0161anai, lai nomain\u012btu objektu atkl\u0101\u0161anu uz to segment\u0101ciju Mask R-CNN.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/L0tzmv--CGY\">YOLO<\/a><\/noindex> (<strong>Y<\/strong>ou <strong>O<\/strong>nly <strong>L<\/strong>ook <strong>O<\/strong>YOLO (You Only Look Once) \u2013 pirm\u0101 neironu t\u012bkls, kas re\u0101llaik\u0101 atpazina objektus mobilaj\u0101s ier\u012bc\u0113s. Iz\u0161\u0137iro\u0161\u0101 iez\u012bme: objekti tiek diferenc\u0113ti vien\u0101 pagriezien\u0101 (svar\u012bgi tikai vienu reizi apskat\u012bt). T\u0101tad YOLO arhitekt\u016br\u0101 nav skaidru \"for\" cilpu, k\u0101 rezult\u0101t\u0101 t\u012bkls darbojas \u0101tri. Piem\u0113ram, \u0161\u0101da analo\u0123ija: NumPy oper\u0101cij\u0101s ar matric\u0101m ar\u012b nav skaidru \"for\" cilpu, kuras NumPy realiz\u0113 zem\u0101k\u0101s arhitekt\u016bras l\u012bme\u0146os, izmantojot programm\u0113\u0161anas valodu C. YOLO izmanto iepriek\u0161 noteiktu logu re\u017e\u0123i. Lai vienu un to pa\u0161u objektu noteiktu vair\u0101kk\u0101rt, tiek izmantots logu p\u0101rkl\u0101juma koeficients (IoU, <strong>I<\/strong>krustojums <strong>o<\/strong>savienojums). \u0160\u012b arhitekt\u016bra darbojas pla\u0161\u0101 diapazon\u0101 un ir augsta <strong>U<\/strong>robustuma <noindex><a rel=\"nofollow\" href=\"https:\/\/ru.wikipedia.org\/wiki\/%D0%A0%D0%BE%D0%B1%D0%B0%D1%81%D1%82%D0%BD%D0%BE%D1%81%D1%82%D1%8C\">: modelis var b\u016bt apm\u0101c\u012bts fotogr\u0101fij\u0101s, bet labi darbojas ar\u012b z\u012bm\u0113t\u0101s glezn\u0101s.<\/a><\/noindex>Vienotais<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/P8e-G-Mhx4k\">SSD<\/a><\/noindex> (<strong>S<\/strong>MultiBox <strong>S<\/strong>detektors) \u2013 tiek izmantoti visefekt\u012bv\u0101kie YOLO arhitekt\u016bras \"haiki\" (piem\u0113ram, non-maximum suppression) un pievieno jaunas, lai neironu t\u012bkls str\u0101d\u0101tu \u0101tr\u0101k un prec\u012bz\u0101k. Iz\u0161\u0137iro\u0161\u0101 iez\u012bme: objektu diferenc\u0113\u0161ana vien\u0101 pagriezien\u0101, izmantojot iepriek\u0161 noteiktu logu re\u017e\u0123i (default box) uz att\u0113lu piram\u012bdas. Att\u0113lu piram\u012bda ir kod\u0113ta konvol\u016bcijas tenzoros, veicot sec\u012bgas konvol\u016bcijas un puli\u0123a oper\u0101cijas (vai maksim\u0101l\u0101s p\u0101rvald\u012bbas proces\u0101 telpisk\u0101 dimensija samazin\u0101s). T\u0101d\u0101 veid\u0101 var noteikt gan lielus, gan mazus objektus vien\u0101 t\u012bkla pagriezien\u0101. <strong>D<\/strong>MobileSSD (<\/li>\n<li>NetV2 +<strong>Mobilais<\/strong>) \u2013 divu neironu t\u012bklu arhitekt\u016bru kombin\u0101cija. Pirmais t\u012bkls <strong>SSD<\/strong>MobileNetV2 <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/352804\/\">darbojas \u0101tri un palielina atpaz\u012b\u0161anas precizit\u0101ti. MobileNetV2 tiek izmantots viet\u0101 VGG-16, kas s\u0101kotn\u0113ji tika izmantota<\/a><\/noindex> . Otrais t\u012bkls SSD nosaka objektu atra\u0161an\u0101s vietu att\u0113l\u0101. <noindex><a rel=\"nofollow\" href=\"https:\/\/arxiv.org\/abs\/1512.02325\">ori\u0123in\u0101laj\u0101 rakst\u0101<\/a><\/noindex>SqueezeNet<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/ge_RT5wvHvY\">\u2013 \u013coti mazs, bet prec\u012bzs neironu t\u012bkls. Tas pats par sevi nesp\u0113j atrisin\u0101t objektu atkl\u0101\u0161anas uzdevumu. Tom\u0113r var tikt izmantots, kombin\u0113jot da\u017e\u0101das arhitekt\u016bras. Un lietojams mobilaj\u0101s ier\u012bc\u0113s. Iz\u0161\u0137iro\u0161\u0101 iez\u012bme ir t\u0101, ka vispirms dati tiek saspiesti l\u012bdz \u010detriem 1\u00d71 konvol\u016bcijas filtriem, un p\u0113c tam papla\u0161in\u0101ti l\u012bdz \u010detriem 1\u00d71 un \u010detriem 3\u00d73 konvol\u016bcijas filtriem. \u0160\u0101da veida datu saspie\u0161anas-papla\u0161in\u0101\u0161anas iter\u0101cija tiek d\u0113v\u0113ta par \"Fire Module\".<\/a><\/noindex> DeepLab<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/b6jhopSMit8\">DeepLab<\/a><\/noindex> (Semantic Image Segmentation with Deep Convolutional Nets) \u2013 objektu segment\u0101cija att\u0113l\u0101. Rakstur\u012bga architekt\u016bras iez\u012bme ir izsviesta (dilated convolution) konvol\u016bcija, kas saglab\u0101 telpisko iz\u0161\u0137irtsp\u0113ju. P\u0113c tam seko rezult\u0101tu p\u0113capstr\u0101des posms, izmantojot grafisko varb\u016bt\u012bbas modeli (conditional random field), kas \u013cauj no\u0146emt nelielus troksn\u012b\u0161us segment\u0101cij\u0101 un uzlabot att\u0113la kvalit\u0101ti. Zem draud\u012bg\u0101 nosaukuma \u00abgrafiskais varb\u016bt\u012bbas modelis\u00bb sl\u0113pjas parasts Gausa filtrs, kas ir aptuveni apr\u0113\u0137in\u0101ts piec\u0101s punktos.<\/li>\n<li>Centos saprast ier\u012bci <noindex><a rel=\"nofollow\" href=\"https:\/\/arxiv.org\/abs\/1711.06897\">RefineDet<\/a><\/noindex> (Single-Shot <strong>Refine<\/strong>ment Neural Network for Object <strong>Det<\/strong>ection), bet maz ko sapratu.<\/li>\n<li>T\u0101pat ar\u012b skat\u012bjos, k\u0101 darbojas \u00abuzman\u012bbas\u00bb (attention) tehnolo\u0123ija: <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/W2rWgXJBZhU\">video1<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/iDulhoQ2pro\">video2<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/H6Qiegq_36c\">video3<\/a><\/noindex>. Rakstur\u012bga arhitekt\u016bras \u00abuzman\u012bba\u00bb iez\u012bme ir autom\u0101tiska paaugstin\u0101t\u0101s uzman\u012bbas re\u0123ionu izcel\u0161ana att\u0113l\u0101 (RoI, <strong>R<\/strong>egions <strong>o<\/strong>f <strong>I<\/strong>nterest) ar neironu t\u012bkla, kas saucas Attention Unit. Paaugstin\u0101tas uzman\u012bbas re\u0123ioni l\u012bdz\u012bgi ierobe\u017eotiem re\u0123ioniem (bounding boxes), ta\u010du at\u0161\u0137ir\u012bb\u0101 no tiem nav fiks\u0113ti att\u0113l\u0101 un var b\u016bt miglaini. P\u0113c tam no paaugstin\u0101tas uzman\u012bbas re\u0123ioniem tika izcelti iez\u012bmes (fichas), kuras \u00abieg\u016bst\u00bb atk\u0101rtojamiem neironu t\u012bkliem ar arhitekt\u016br\u0101m <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/5lUUrREboSk\">LSDM, GRU vai Vanilla RNN<\/a><\/noindex>. Atk\u0101rtojamie neironu t\u012bkli sp\u0113j analiz\u0113t attiec\u012bbas iez\u012bm\u0113s sec\u012bb\u0101. Atk\u0101rtojamie neironu t\u012bkli s\u0101kotn\u0113ji tika izmantoti tekstu tulko\u0161anai cit\u0101s valod\u0101s, bet tagad ar\u012b att\u0113la tulko\u0161anai <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">no att\u0113la uz tekstu<\/a><\/noindex> un <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">no teksta uz att\u0113lu<\/a><\/noindex>.<\/li>\n<\/ul>\n<p><\/p>\n<p>P\u0113tot \u0161\u012bs arhitekt\u016bras <strong>es sapratu, ka neko nesaprotu<\/strong>. Un jaut\u0101jums nav par to, ka manam neironu t\u012bklam ir probl\u0113mas ar uzman\u012bbas meh\u0101nismu. \u0160o visu arhitekt\u016bru rad\u012b\u0161ana l\u012bdzin\u0101s kaut k\u0101dam milz\u012bgam hackathon, kur autori sacen\u0161as hacko\u0161an\u0101. Hacks (hack) \u2013 \u0101trs risin\u0101jums gr\u016btai programmat\u016bras probl\u0113mai. Tas ir, starp vis\u0101m \u0161\u012bm arhitekt\u016br\u0101m nav redzamas un saprotamas lo\u0123iskas saiknes. Viss, kas t\u0101s vieno \u2013 ir visveiksm\u012bg\u0101ko hacks kopums, ko t\u0101s aiz\u0146emas viena no otras, plus kop\u012bga vis\u0101m <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/Ilg3gGewQ5U\">konvol\u016bcijas oper\u0101cija ar atgriezenisko saiti<\/a><\/noindex> (atgriezenisk\u0101 izplat\u012b\u0161ana, backpropagation). Nav <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/272473\/\">sist\u0113misk\u0101 dom\u0101\u0161ana<\/a><\/noindex>! Nav skaidrs, ko main\u012bt un k\u0101 optimiz\u0113t eso\u0161os sasniegumus.<\/p>\n<p><\/p>\n<p>K\u0101 rezult\u0101ts lo\u0123isk\u0101s saiknes tr\u016bkumam starp h\u0101kiem, tos ir \u0101rk\u0101rt\u012bgi gr\u016bti atcer\u0113ties un pielietot praks\u0113. Tas ir fragment\u0113ts zin\u0101\u0161anas. Lab\u0101kaj\u0101 gad\u012bjum\u0101 tiek iegaum\u0113ti da\u017ei interesanti un negaid\u012bti momenti, bet liel\u0101k\u0101 da\u013ca no saprast\u0101 un nesaprast\u0101 izn\u012bkst atmi\u0146\u0101 jau p\u0113c da\u017e\u0101m dien\u0101m. B\u016btu labi, ja p\u0113c ned\u0113\u013cas atcer\u0113tos vismaz arhitekt\u016bras nosaukumu. Bet rakstu las\u012b\u0161anai un apskatu video skat\u012b\u0161anai tika pat\u0113r\u0113tas stundas un pat dienas darba laika!<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neironu t\u012bkli. Kur viss tas virz\u0101s\" src=\"\/wp-content\/uploads\/2020\/01\/86f0f24e3be1d0f1a210f3132897d981.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Att\u0113ls 2 \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/www.asimovinstitute.org\/neural-network-zoo\/\">Neironu t\u012bklu zoolo\u0123iskais d\u0101rzs<\/a><\/noindex><\/p>\n<p><\/p>\n<p>Vairums zin\u0101tnisko rakstu autoru, manupr\u0101t, dara visu iesp\u0113jamo, lai pat \u0161\u012bs fragment\u0113t\u0101s zin\u0101\u0161anas netiktu saprastas no las\u012bt\u0101ja. Bet darb\u012bbas v\u0101rdu formas desmit rind\u0101s teikumos ar formul\u0101m, kas \u0146emtas \u2018no griestiem\u2019 \u2013 tas ir atsevi\u0161\u0137as raksta t\u0113ma (probl\u0113ma <noindex><a rel=\"nofollow\" href=\"https:\/\/en.wikipedia.org\/wiki\/Publish_or_perish\">publish or perish<\/a><\/noindex>).<\/p>\n<p><\/p>\n<p>T\u0101p\u0113c rad\u0101s nepiecie\u0161am\u012bba sistematiz\u0113t inform\u0101ciju par neironu t\u012bkliem un t\u0101d\u0113j\u0101di uzlabot izpratnes un atcer\u0113\u0161an\u0101s kvalit\u0101ti. T\u0101p\u0113c galven\u0101 t\u0113ma atsevi\u0161\u0137o tehnolo\u0123iju un m\u0101ksl\u012bgo neironu t\u012bklu arhitekt\u016bru izp\u0113t\u0113 ir k\u013cuvusi \u0161\u0101da uzdevums: <strong>u zakot, kur tas viss virz\u0101s<\/strong>, nevis k\u0101da konkr\u0113ta neironu t\u012bkla uzb\u016bve.<\/p>\n<p><\/p>\n<p>Kur viss tas virz\u0101s. Galvenie rezult\u0101ti:<\/p>\n<p><\/p>\n<ul>\n<li>S\u0101kuma skaits ma\u0161\u012bnm\u0101c\u012bbas uz\u0146\u0113mumos p\u0113d\u0113jo divu gadu laik\u0101 <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/recognitor\/blog\/455676\/\">strauji samazin\u0101j\u0101s<\/a><\/noindex>. Iesp\u0113jam\u0101 iemesls: \u2018neironu t\u012bkli vairs nav kas jauns\u2019.<\/li>\n<li>Ikviens sp\u0113s izveidot funkcion\u0113jo\u0161u neironu t\u012bklu vienk\u0101r\u0161ai uzdevuma risin\u0101\u0161anai. Lai to izdar\u012btu, vi\u0146\u0161 pa\u0146ems gatavu modeli no \u2018mode\u013cu zoolo\u0123isk\u0101 d\u0101rza\u2019 (model zoo) un apm\u0101c\u012bs p\u0113d\u0113jo neironu t\u012bkla sl\u0101ni (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/yofjFQddwHE\">transfer learning<\/a><\/noindex>) uz gataviem datiem no <noindex><a rel=\"nofollow\" href=\"https:\/\/toolbox.google.com\/datasetsearch\">Google Dataset Search<\/a><\/noindex> vai no <noindex><a rel=\"nofollow\" href=\"https:\/\/www.kaggle.com\/datasets\">25 t\u016bksto\u0161iem Kaggle datu kop\u0101m<\/a><\/noindex> bez maksas <noindex><a rel=\"nofollow\" href=\"https:\/\/www.dataschool.io\/cloud-services-for-jupyter-notebook\/\">Jupyter Notebook m\u0101ko\u0146os<\/a><\/noindex>.<\/li>\n<li>Lieli neironu t\u012bklu ra\u017eot\u0101ji s\u0101ka veidot <strong>\u2018mode\u013cu zoolo\u0123iskos d\u0101rzus\u2019<\/strong> (model zoo). Ar to pal\u012bdz\u012bbu var \u0101tri izveidot komerci\u0101lu lietojumprogrammu: <noindex><a rel=\"nofollow\" href=\"https:\/\/tfhub.dev\/\">TF Hub<\/a><\/noindex> TensorFlow, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/open-mmlab\/mmdetection\">MMDetection<\/a><\/noindex> PyTorch, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/facebookresearch\/Detectron\">Detectron<\/a><\/noindex> Caffe2, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/wkentaro\/chainer-modelzoo\">chainer-modelzoo<\/a><\/noindex> Chainer un <noindex><a rel=\"nofollow\" href=\"https:\/\/modelzoo.co\/\">citi<\/a><\/noindex>.<\/li>\n<li>Neironu t\u012bkli, kas darbojas <strong>re\u0101l\u0101 laik\u0101<\/strong> (real-time) mobilaj\u0101s ier\u012bc\u0113s. No 10 l\u012bdz 50 kadriem sekund\u0113.<\/li>\n<li>Neironu t\u012bklu pielietojums telefonos (TF Lite), p\u0101rl\u016bkos (TF.js) un <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/19ZNz2N79u4\">ikdienas priek\u0161metos<\/a><\/noindex> (IoT, <strong>I<\/strong>interneta <strong>o<\/strong>f <strong>T<\/strong>lietas). \u012apa\u0161i telefonos, kuri jau atbalsta neironu t\u012bklus uz \u2018dzelzs\u2019 l\u012bme\u0146a (neiroakseleratori).<\/li>\n<li>\u201eKatram ier\u012bc\u0113m, ap\u0123\u0113rba priek\u0161metiem un iesp\u0113jams pat p\u0101rtikai b\u016bs <strong>IP-v6 adrese<\/strong> un t\u0101s sazin\u0101sies sav\u0101 starp\u0101\u201c \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/GG7H8Xa4m8I?t=85\">Sebastians Truns<\/a><\/noindex>.<\/li>\n<li>Arvien pieaug public\u0113jumu skaits ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s jom\u0101, kas s\u0101ka <noindex><a rel=\"nofollow\" href=\"http:\/\/data-mining.philippe-fournier-viger.com\/too-many-machine-learning-papers\">p\u0101rsniegt Moora likumu<\/a><\/noindex> (dubultojas reizi divos gados) no 2015. gada. Ac\u012bmredzot ir nepiecie\u0161amas neironu t\u012bklos balst\u012btas sist\u0113mas rakstu anal\u012bzei.<\/li>\n<li>Popul\u0101ras k\u013c\u016bst \u0161\u0101das tehnolo\u0123ijas:\n<ul>\n<li><strong>PyTorch<\/strong> \u2013 popularit\u0101te strauji pieaug un, \u0161\u0137iet, apsteidz TensorFlow.<\/li>\n<li>Autom\u0101tiska hiperp\u0101rparametru izv\u0113le <strong>AutoML<\/strong> \u2013 popularit\u0101te pieaug pak\u0101peniski.<\/li>\n<li>Pak\u0101peniska precizit\u0101tes samazin\u0101\u0161an\u0101s un izpildes \u0101truma palielin\u0101\u0161an\u0101s: <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rln_kZbYaWc\">neprec\u012bza lo\u0123ika<\/a><\/noindex>, algoritmi <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/MIPkK5ZAsms\">veicin\u0101\u0161anai<\/a><\/noindex>, neprec\u012bza (sal\u012bdzino\u0161a) apr\u0113\u0137in\u0101\u0161ana, kvantiz\u0101cija (kad neironu t\u012bkla svars tiek p\u0101rveidots par veseliem skait\u013ciem un kvantiz\u0113ts), neironu akseleratori.<\/li>\n<li>Tulkojums <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">no att\u0113la uz tekstu<\/a><\/noindex> un <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">no teksta uz att\u0113lu<\/a><\/noindex>.<\/li>\n<li>Izveide <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/OrHLacCDZVQ\">tridimension\u0101lu objektu anal\u012bze video<\/a><\/noindex>, tagad jau re\u0101l\u0101 laik\u0101.<\/li>\n<li>Galvenais DL \u2013 ir daudz datu, bet tos sav\u0101kt un anot\u0113t nav viegli. T\u0101p\u0113c att\u012bst\u0101s autom\u0101tisk\u0101 anot\u0101cija (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/NcKTn4C91Yc\">automated annotation<\/a><\/noindex>) neironu t\u012bklu lieto\u0161anai, izmantojot neironu t\u012bklus.<\/li>\n<\/ul>\n<\/li>\n<li>Ar neironu t\u012bkliem datorzin\u0101tne p\u0113k\u0161\u0146i k\u013cuvusi <strong>eksperiments zin\u0101tne<\/strong> un rad\u0101s <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/480348\">reproduc\u0113jam\u012bbas kr\u012bze<\/a><\/noindex>.<\/li>\n<li>IT nauda un neironu t\u012bklu popularit\u0101te rad\u0101s vienlaikus, kad apr\u0113\u0137ini k\u013cuva par tirgus v\u0113rt\u012bbu. Ekonomika no zeltu un val\u016btu k\u013c\u016bst <strong>zelta-vice-apr\u0113\u0137in\u0101\u0161anas<\/strong>. Skatiet manu rakstu par <noindex><a rel=\"nofollow\" href=\"https:\/\/ru.wikipedia.org\/wiki\/%D0%AD%D0%BA%D0%BE%D0%BD%D0%BE%D1%84%D0%B8%D0%B7%D0%B8%D0%BA%D0%B0\">ekonofiziku<\/a><\/noindex> un IT naudas ra\u0161an\u0101s iemeslu.<\/li>\n<\/ul>\n<p><\/p>\n<p>Pak\u0101peniski par\u0101d\u0101s jauna <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/481844\">ML\/DL programm\u0113\u0161anas metodolo\u0123ija<\/a><\/noindex> (Ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s un dzi\u013c\u0101 m\u0101c\u012b\u0161an\u0101s), kas balst\u0101s uz programmas izpratni k\u0101 apm\u0101c\u012btu neironu t\u012bklu mode\u013cu kopumu.<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neironu t\u012bkli. Kur viss tas virz\u0101s\" src=\"\/wp-content\/uploads\/2020\/01\/17df76a613191522145abeba177c2685.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Att\u0113ls 3 \u2013 ML\/DL k\u0101 jauna programm\u0113\u0161anas metodolo\u0123ija<\/p>\n<p><\/p>\n<p>Tom\u0113r nav izveidojusies <strong>\u201eneironu t\u012bklu teorija\u201d<\/strong>, kas \u013cautu dom\u0101t un str\u0101d\u0101t sistem\u0101tiski. Tas, ko \u0161obr\u012bd sauc par \u201eteoriju\u201d, paties\u012bb\u0101 ir eksperiment\u0101li, heoristiski algoritmi.<\/p>\n<p><\/p>\n<p>Saites uz maniem un ne tikai resursiem:<\/p>\n<p><\/p>\n<ul>\n<li>Jaunumu bi\u013cetens par Datu zin\u0101tni. Galvenok\u0101rt par att\u0113lu apstr\u0101di. Tie, kas v\u0113las sa\u0146emt, l\u016bdzu, s\u016btiet e-pastu (foobar167&lt;gaff-gaff&gt;gmail&lt;punktu&gt;com). Saite uz rakstiem un video tiek nos\u016bt\u012btas, kad materi\u0101ls tiek uzkr\u0101ts.<\/li>\n<li>Visp\u0101r\u0113js <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Machine_Learning\/courses_on_machine_learning.md\">kursu un rakstu saraksts<\/a><\/noindex>, kurus esmu izg\u0101jis un kurus v\u0113l\u0113tos apg\u016bt.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Ubuntu\/13_Keras_and_TensorFlow_how-tos.md#exercises\">Kursi un video ies\u0101c\u0113jiem<\/a><\/noindex>, no kuriem b\u016btu j\u0101s\u0101k m\u0101c\u012bties par neironu t\u012bkliem. Plus bro\u0161\u016bra <noindex><a rel=\"nofollow\" href=\"https:\/\/foobar167.github.io\/page\/vvedeniye-v-mashinnoye-obucheniye-i-iskusstvennyye-neyronnyye-seti.html\">\u201eIevads ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101 un m\u0101ksl\u012bgajos neironu t\u012bklos\u201d<\/a><\/noindex>.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Ubuntu\/13_Keras_and_TensorFlow_how-tos.md#tools\">Noder\u012bgi r\u012bki<\/a><\/noindex>, kur katrs atrad\u012bs kaut ko interesantu sev.<\/li>\n<li>Ekstrem\u0101li noder\u012bgi ir biju\u0161i <strong>video kan\u0101li, kas analiz\u0113 zin\u0101tniskos rakstus<\/strong> par Data Science. Atrodiet, abon\u0113jiet tos un dalieties ar sait\u0113m ar saviem kol\u0113\u0123iem un ar mani ar\u012b. Piem\u0113ri:\n<ul>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/user\/keeroyz\">Two Minute Papers<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/channel\/UCHB9VepY6kYvZjj0Bgxnpbw\">Henry AI Labs<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/channel\/UCZHmQk67mSJgfCCTn7xBfew\">Yannic Kilcher<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/channel\/UC5_6ZD6s8klmMu9TXEB_1IA\">CodeEmporium<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.dlology.com\">Chengwei Zhang blogs<\/a><\/noindex> un <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/Tony607\">Tony607<\/a><\/noindex> ar soli pa solim instrukcij\u0101m un atv\u0113rt\u0101 koda projektiem.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p><\/p>\n<p>Paldies par uzman\u012bbu!<\/p>\n<p>Avots: <a content=\"nofollow\" rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/482794\/\">habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"<p>\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b. \u0421\u0442\u0430\u0440\u0430\u043b\u0441\u044f \u0432\u044b\u0431\u0438\u0440\u0430\u0442\u044c \u0432\u0438\u0434\u0435\u043e \u043f\u043e\u044f\u0441\u043d\u0435\u043d\u0438\u044f \u0438 \u0436\u0435\u043b\u0430\u0442\u0435\u043b\u044c\u043d\u043e \u043d\u0430 \u0440\u0443\u0441\u0441\u043a\u043e\u043c \u044f\u0437\u044b\u043a\u0435. \u0412\u0442\u043e\u0440\u0430\u044f \u0447\u0430\u0441\u0442\u044c \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0432 \u043f\u043e\u043f\u044b\u0442\u043a\u0435 \u043e\u0441\u043e\u0437\u043d\u0430\u0442\u044c \u043d\u0430\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435 \u0440\u0430\u0437\u0432\u0438\u0442\u0438\u044f \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u043d\u0435\u0439\u0440\u043e\u043d\u043d\u044b\u0445 \u0441\u0435\u0442\u0435\u0439. \u0418 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0439 \u043d\u0430 \u0438\u0445 \u043e\u0441\u043d\u043e\u0432\u0435. \u0420\u0438\u0441\u0443\u043d\u043e\u043a 1 \u2013 \u041f\u043e\u043d\u0438\u043c\u0430\u0442\u044c [&hellip;]<\/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-54780","post","type-post","status-publish","format-standard","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t<!-- All in One SEO 5.0.5 - aioseo.com -->\n\t<meta name=\"description\" content=\"\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b.\" \/>\n\t<meta name=\"robots\" content=\"max-image-preview:large\" \/>\n\t<meta name=\"author\" content=\"Yuri Gagarin\"\/>\n\t<link rel=\"canonical\" href=\"https:\/\/prohoster.info\/lv\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya\" \/>\n\t\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.5\" \/>\n\t\t<meta property=\"og:locale\" content=\"lv_LV\" \/>\n\t\t<meta property=\"og:site_name\" content=\"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\" \/>\n\t\t<meta property=\"og:type\" content=\"article\" \/>\n\t\t<meta property=\"og:title\" content=\"\ud83e\udd47\u041d\u0435\u0439\u0440\u043e\u0441\u0435\u0442\u0438. \u041a\u0443\u0434\u0430 \u044d\u0442\u043e \u0432\u0441\u0435 \u0434\u0432\u0438\u0436\u0435\u0442\u0441\u044f | ProHoster\" \/>\n\t\t<meta property=\"og:description\" content=\"\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b.\" \/>\n\t\t<meta property=\"og:url\" content=\"https:\/\/prohoster.info\/lv\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya\" \/>\n\t\t<meta property=\"og:image\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:secure_url\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:width\" content=\"350\" \/>\n\t\t<meta property=\"og:image:height\" content=\"350\" \/>\n\t\t<meta property=\"article:published_time\" content=\"2020-01-03T21:00:00+00:00\" \/>\n\t\t<meta property=\"article:modified_time\" content=\"2020-02-18T11:02:50+00:00\" \/>\n\t\t<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<!-- All in One SEO -->\n\n","aioseo_head_json":{"title":"\ud83e\udd47Neironu t\u012bkli. Uz kurieni tas viss virz\u0101s | ProHoster","description":"Raksts sast\u0101v no div\u0101m da\u013c\u0101m: \u012bss apraksts par da\u017e\u0101d\u0101m t\u012bklu arhitekt\u016br\u0101m objektu atkl\u0101\u0161anai att\u0113los un att\u0113lu segment\u0113\u0161anai ar visvienk\u0101r\u0161\u0101kajiem man saprotamajiem saist\u0113m uz resursiem.","canonical_url":"https:\/\/prohoster.info\/lv\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"lv_LV","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\udd47\u041d\u0435\u0439\u0440\u043e\u0441\u0435\u0442\u0438. \u041a\u0443\u0434\u0430 \u044d\u0442\u043e \u0432\u0441\u0435 \u0434\u0432\u0438\u0436\u0435\u0442\u0441\u044f | ProHoster","og:description":"\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b.","og:url":"https:\/\/prohoster.info\/lv\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","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":"2020-01-03T21:00:00+00:00","article:modified_time":"2020-02-18T11:02:50+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"54780","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":"2026-01-24 12:43:20","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-02-28 19:57:38","updated":"2026-01-24 12:43:20","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/posts\/54780","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/comments?post=54780"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/posts\/54780\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/media?parent=54780"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/categories?post=54780"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/tags?post=54780"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}