{"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\/sk\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","title":{"rendered":"Neur\u00f3nov\u00e9 siete. Kam to v\u0161etko smeruje","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>\u010cl\u00e1nok sa sklad\u00e1 z dvoch \u010dast\u00ed:<\/p>\n<p><\/p>\n<ol>\n<li>Kr\u00e1tky popis niektor\u00fdch architekt\u00far siet\u00ed na detekciu objektov na obr\u00e1zkoch a segment\u00e1ciu obr\u00e1zkov s najzrozumite\u013enej\u0161\u00edmi odkazmi na zdroje. Sna\u017eil som sa vybera\u0165 video vysvetlenia a ide\u00e1lne v ru\u0161tine.<\/li>\n<li>Druh\u00e1 \u010das\u0165 spo\u010d\u00edva v pokuse pochopi\u0165 smer v\u00fdvoja architekt\u00far neur\u00f3nov\u00fdch siet\u00ed a technol\u00f3gi\u00ed na ich z\u00e1klade.<\/li>\n<\/ol>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neur\u00f3nov\u00e9 siete. Kam to v\u0161etko smeruje\" src=\"\/wp-content\/uploads\/2020\/01\/3e0238547dc956dbb069c11241e4534f.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Obr\u00e1zok 1 \u2013 Pochopi\u0165 architekt\u00fary neur\u00f3nov\u00fdch siet\u00ed nie je jednoduch\u00e9<\/p>\n<p><\/p>\n<p>V\u0161etko za\u010dalo t\u00fdm, \u017ee som vytvoril dve demon\u0161tra\u010dn\u00e9 aplik\u00e1cie na klasifik\u00e1ciu a detekciu objektov na telef\u00f3ne Android:<\/p>\n<p><\/p>\n<ul>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/junkyard\/tree\/master\/object_classifier\">Back-end demo<\/a><\/noindex>, ke\u010f sa \u00fadaje sprac\u00favaj\u00fa na serveri a pren\u00e1\u0161aj\u00fa na telef\u00f3n. Klasifik\u00e1cia obr\u00e1zkov (image classification) troch typov medve\u010fov: hned\u00e9ho, \u010dierneho a ply\u0161ov\u00e9ho.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/android\/tree\/master\/object_detection_demo\">Front-end demo<\/a><\/noindex>, ke\u010f sa \u00fadaje sprac\u00favaj\u00fa priamo na telef\u00f3ne. Detekcia objektov (object detection) troch typov: lieskov\u00fd orech, figy a datle.<\/li>\n<\/ul>\n<p><noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex><\/p>\n<p>Existuje rozdiel medzi \u00falohami klasifik\u00e1cie obr\u00e1zkov, detekciou objektov na obr\u00e1zku a <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/analytics-vidhya\/image-classification-vs-object-detection-vs-image-segmentation-f36db85fe81\">segment\u00e1ciou obr\u00e1zkov<\/a><\/noindex>. Preto vznikla potreba zisti\u0165, ktor\u00e9 architekt\u00fary neur\u00f3nov\u00fdch siet\u00ed deteguj\u00fa objekty na obr\u00e1zkoch a ktor\u00e9 m\u00f4\u017eu segmentova\u0165. Na\u0161iel som nasleduj\u00face pr\u00edklady architekt\u00far s najzrozumite\u013enej\u0161\u00edmi odkazmi na zdroje:<\/p>\n<p><\/p>\n<ul>\n<li>S\u00e9ria architekt\u00far na b\u00e1ze R-CNN (<strong>R<\/strong>regi\u00f3ny s <strong>C<\/strong>konvolu\u010dn\u00fdmi <strong>N<\/strong>neur\u00f3nov\u00fdmi <strong>N<\/strong>sie\u0165ami): 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>. Na detekciu objektu na obr\u00e1zku sa pomocou mechanizmu Region Proposal Network (RPN) vyberaj\u00fa obmedzen\u00e9 regi\u00f3ny (bounding boxes). P\u00f4vodne sa namiesto RPN pou\u017e\u00edval pomal\u0161\u00ed mechanizmus Selective Search. Potom vybran\u00e9 obmedzen\u00e9 regi\u00f3ny sa posielaj\u00fa na vstup be\u017enej neur\u00f3novej siete na klasifik\u00e1ciu. V architekt\u00fare R-CNN s\u00fa jasn\u00e9 cykly \u201efor\u201c prech\u00e1dzania cez obmedzen\u00e9 regi\u00f3ny, celkovo a\u017e 2000 prechodov cez intern\u00fa sie\u0165 AlexNet. V d\u00f4sledku jasn\u00fdch cyklov \u201efor\u201c sa spoma\u013euje r\u00fdchlos\u0165 spracovania obr\u00e1zkov. Po\u010det jasn\u00fdch cyklov, prechodov cez intern\u00fa neur\u00f3nov\u00fa sie\u0165, sa zni\u017euje s ka\u017edou novou verziou architekt\u00fary, pri\u010dom sa vykon\u00e1va desiatky \u010fal\u0161\u00edch zmien na zv\u00fd\u0161enie r\u00fdchlosti a na nahradenie \u00falohy detekcie objektov segment\u00e1ciou objektov v Mask R-CNN.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/L0tzmv--CGY\">YOLO<\/a><\/noindex> (<strong>Y<\/strong>len <strong>O<\/strong>pozrite <strong>L<\/strong>kniha <strong>O<\/strong>nce) \u2013 prv\u00e1 neur\u00f3nov\u00e1 sie\u0165, ktor\u00e1 rozpozn\u00e1vala objekty v re\u00e1lnom \u010dase na mobiln\u00fdch zariadeniach. Vyzna\u010duje sa schopnos\u0165ou rozli\u0161ova\u0165 objekty na jeden prechod (sta\u010d\u00ed sa pozrie\u0165 raz). To znamen\u00e1, \u017ee v architekt\u00fare YOLO nie s\u00fa \u017eiadne jasn\u00e9 cykly \u201efor\u201c, \u010do umo\u017e\u0148uje r\u00fdchlu pr\u00e1cu siete. Napr\u00edklad tak\u00e1 anal\u00f3gia: pri oper\u00e1ci\u00e1ch s maticami v NumPy nie s\u00fa taktie\u017e \u017eiadne jasn\u00e9 cykly \u201efor\u201c, ktor\u00e9 s\u00fa realizovan\u00e9 na ni\u017e\u0161\u00edch \u00farovniach architekt\u00fary prostredn\u00edctvom programovacieho jazyka C. YOLO pou\u017e\u00edva mrie\u017eku z vopred ur\u010den\u00fdch okien. Aby sa ten ist\u00fd objekt neidentifikoval viackr\u00e1t, pou\u017e\u00edva sa koeficient prekrytia okien (IoU, <strong>I<\/strong>ntersection <strong>o<\/strong>ver <strong>U<\/strong>nion). T\u00e1to architekt\u00fara funguje v \u0161irokom rozsahu a m\u00e1 vysok\u00fa <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\">robustnos\u0165<\/a><\/noindex>: model m\u00f4\u017ee by\u0165 tr\u00e9novan\u00fd na fotografi\u00e1ch, ale pritom dobre fungova\u0165 aj na kreslen\u00fdch obrazcoch.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/P8e-G-Mhx4k\">SSD<\/a><\/noindex> (<strong>S<\/strong>ingle <strong>S<\/strong>hot MultiBox <strong>D<\/strong>etector) \u2013 vyu\u017e\u00edva naj\u00faspe\u0161nej\u0161ie \u201ehacky\u201c architekt\u00fary YOLO (napr\u00edklad non-maximum suppression) a prid\u00e1va nov\u00e9, aby neurov\u00e1 sie\u0165 pracovala r\u00fdchlej\u0161ie a presnej\u0161ie. Vyzna\u010duje sa schopnos\u0165ou rozli\u0161ova\u0165 objekty na jeden prechod za pomoci vopred nastavenej mrie\u017eky okien (default box) na pyram\u00eddach obr\u00e1zkov. Pyram\u00edda obr\u00e1zkov je zak\u00f3dovan\u00e1 v konvolu\u010dn\u00fdch tenzoroch pri sekven\u010dn\u00fdch oper\u00e1ci\u00e1ch konvol\u00facie a pooling. (pri oper\u00e1cii max-pooling sa priestorov\u00e1 dimenzia zmen\u0161uje). T\u00fdmto sp\u00f4sobom s\u00fa rozpozn\u00e1van\u00e9 ve\u013ek\u00e9 aj mal\u00e9 objekty na jeden prechod siete.<\/li>\n<li>MobileSSD (<strong>Mobiln\u00e9<\/strong>NetV2 + <strong>SSD<\/strong>) \u2013 kombin\u00e1cia dvoch architekt\u00far neur\u00f3nov\u00fdch siet\u00ed. Prv\u00e1 sie\u0165 <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/352804\/\">MobileNetV2<\/a><\/noindex> pracuje r\u00fdchlo a zvy\u0161uje presnos\u0165 rozpozn\u00e1vania. MobileNetV2 sa pou\u017e\u00edva namiesto VGG-16, ktor\u00e1 bola p\u00f4vodne pou\u017e\u00edvan\u00e1 v <noindex><a rel=\"nofollow\" href=\"https:\/\/arxiv.org\/abs\/1512.02325\">origin\u00e1lnom \u010dl\u00e1nku<\/a><\/noindex>. Druh\u00e1 sie\u0165 SSD ur\u010duje polohu objektov na obr\u00e1zku.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/ge_RT5wvHvY\">SqueezeNet<\/a><\/noindex> \u2013 ve\u013emi mal\u00e1, ale presn\u00e1 neur\u00f3nov\u00e1 sie\u0165. Sama o sebe nerie\u0161i probl\u00e9m detekcie objektov. Av\u0161ak m\u00f4\u017ee by\u0165 pou\u017eit\u00e1 v kombin\u00e1cii r\u00f4znych architekt\u00far. A pou\u017e\u00edva\u0165 sa m\u00f4\u017ee v mobiln\u00fdch zariadeniach. Vyzna\u010duje sa t\u00fdm, \u017ee najprv s\u00fa d\u00e1ta komprimovan\u00e9 na \u0161tyri 1\u00d71 konvolu\u010dn\u00e9 filtre a potom roz\u0161\u00edren\u00e9 na \u0161tyri 1\u00d71 a \u0161tyri 3\u00d73 konvolu\u010dn\u00e9 filtre. Jedna tak\u00e1 iter\u00e1cia kompresie-roz\u0161\u00edrenia d\u00e1t sa naz\u00fdva \u201eFire Module\u201c.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/b6jhopSMit8\">DeepLab<\/a><\/noindex> (Semantic Image Segmentation with Deep Convolutional Nets) \u2013 segment\u00e1cia objektov na obr\u00e1zku. Hlavnou \u010drtou architekt\u00fary je rozpt\u00fdlen\u00e1 (dilated convolution) konvol\u00facia, ktor\u00e1 zachov\u00e1va priestorov\u00e9 rozl\u00ed\u0161enie. Nasleduje f\u00e1za postprocessing v\u00fdsledkov s pou\u017eit\u00edm grafick\u00e9ho pravdepodobnostn\u00e9ho modelu (conditional random field), \u010do umo\u017e\u0148uje odstr\u00e1ni\u0165 mal\u00e9 \u0161umy v segment\u00e1cii a zlep\u0161i\u0165 kvalitu segmentovan\u00e9ho obr\u00e1zka. Pod hroziv\u00fdm n\u00e1zvom \u201egrafick\u00fd pravdepodobnostn\u00fd model\u201c sa skr\u00fdva be\u017en\u00fd Gaussov filter, ktor\u00fd je aproximovan\u00fd pod\u013ea piatich bodov.<\/li>\n<li>Sna\u017eil som sa pochopi\u0165 zariadenie <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), ale m\u00e1lo som pochopil.<\/li>\n<li>Tie\u017e som sa pozrel, ako funguje technol\u00f3gia \u201epozornos\u0165\u201c (attention): <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>. Hlavnou \u010drtou architekt\u00fary \u201epozornos\u0165\u201c je automatick\u00e9 vy\u010dlenenie regi\u00f3nov zv\u00fd\u0161enej pozornosti na obr\u00e1zku (RoI, <strong>R<\/strong>egions <strong>o<\/strong>f <strong>I<\/strong>nterest) pomocou neur\u00f3novej siete nazvanej Attention Unit. Regi\u00f3ny zv\u00fd\u0161enej pozornosti s\u00fa podobn\u00e9 obmedzen\u00fdm regi\u00f3nom (bounding boxes), ale na rozdiel od nich nie s\u00fa fixovan\u00e9 na obr\u00e1zku a m\u00f4\u017eu ma\u0165 rozmazan\u00e9 hranice. N\u00e1sledne sa z regi\u00f3nov zv\u00fd\u0161enej pozornosti vyberaj\u00fa znaky (features), ktor\u00e9 sa \u201enap\u00e1jaj\u00fa\u201c do rekurentn\u00fdch neur\u00f3nov\u00fdch siet\u00ed s architekt\u00farami <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/5lUUrREboSk\">LSDM, GRU alebo Vanilla RNN<\/a><\/noindex>. Rekurentn\u00e9 neur\u00f3nov\u00e9 siete dok\u00e1\u017eu analyzova\u0165 vz\u0165ah znakov v sekvencii. Rekurentn\u00e9 neur\u00f3nov\u00e9 siete sa p\u00f4vodne pou\u017e\u00edvali na preklady textu do in\u00fdch jazykov, a teraz aj na preklad <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">obr\u00e1zku do textu<\/a><\/noindex> a <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">textu do obr\u00e1zku<\/a><\/noindex>.<\/li>\n<\/ul>\n<p><\/p>\n<p>Po\u010das \u0161t\u00fadia t\u00fdchto architekt\u00far <strong>som pochopil, \u017ee ni\u010domu nerozumiem<\/strong>. A nejde o to, \u017ee by moja neur\u00f3nov\u00e1 sie\u0165 mala probl\u00e9my s mechanizmom pozornosti. Tvorenie v\u0161etk\u00fdch t\u00fdchto architekt\u00far sa zd\u00e1 ako nejak\u00fd obrovsk\u00fd hackathon, kde autori s\u00fa\u0165a\u017eia v hackoch. Hack (hack) \u2013 r\u00fdchle rie\u0161enie \u0165a\u017ekej programovej \u00falohy. To znamen\u00e1, \u017ee medzi v\u0161etk\u00fdmi t\u00fdmito architekt\u00farami nie je vidite\u013en\u00e1 a zrozumite\u013en\u00e1 logick\u00e1 s\u00favislos\u0165. V\u0161etko, \u010do ich sp\u00e1ja \u2013 je s\u00fabor naj\u00faspe\u0161nej\u0161\u00edch hackov, ktor\u00e9 si navz\u00e1jom po\u017ei\u010diavaj\u00fa, plus spolo\u010dn\u00e1 pre v\u0161etk\u00fdch <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/Ilg3gGewQ5U\">oper\u00e1cia konvol\u00facie s sp\u00e4tnou v\u00e4zbou<\/a><\/noindex> (sp\u00e4tn\u00e9 \u0161\u00edrenie chyby, backpropagation). Nie je <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/272473\/\">syst\u00e9mov\u00e9 myslenie<\/a><\/noindex>! Nie je jasn\u00e9, \u010do meni\u0165 a ako optimalizova\u0165 existuj\u00face dosiahnutia.<\/p>\n<p><\/p>\n<p>V d\u00f4sledku absencie logickej v\u00e4zby medzi hackmi je mimoriadne \u0165a\u017ek\u00e9 si ich zapam\u00e4ta\u0165 a prakticky ich uplatni\u0165. Ide o fragmentovan\u00e9 znalosti. V tom najlep\u0161om pr\u00edpade si zapam\u00e4t\u00e1me nieko\u013eko zauj\u00edmav\u00fdch a neo\u010dak\u00e1van\u00fdch momentov, ale v\u00e4\u010d\u0161ina z toho, \u010do bolo pochopen\u00e9 a nepochopen\u00e9, sa u\u017e o nieko\u013eko dn\u00ed vytrat\u00ed z pam\u00e4ti. Bolo by dobr\u00e9, keby sa aspo\u0148 n\u00e1zov architekt\u00fary pripomenul za t\u00fd\u017ede\u0148. A to pritom bolo na \u010d\u00edtanie \u010dl\u00e1nkov a pozeranie preh\u013eadov\u00fdch vide\u00ed vynalo\u017een\u00e9 nieko\u013eko hod\u00edn a dokonca dn\u00ed pracovn\u00e9ho \u010dasu!<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neur\u00f3nov\u00e9 siete. Kam to v\u0161etko smeruje\" src=\"\/wp-content\/uploads\/2020\/01\/86f0f24e3be1d0f1a210f3132897d981.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Obr\u00e1zok 2 \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/www.asimovinstitute.org\/neural-network-zoo\/\">ZOO neur\u00f3nov\u00fdch siet\u00ed<\/a><\/noindex><\/p>\n<p><\/p>\n<p>V\u00e4\u010d\u0161ina autorov vedeck\u00fdch \u010dl\u00e1nkov, pod\u013ea m\u00f4jho osobn\u00e9ho n\u00e1zoru, rob\u00ed v\u0161etko pre to, aby aj tieto fragmentovan\u00e9 znalosti neboli \u010ditate\u013eom pochopen\u00e9. Ale gerund\u00edvne kon\u0161trukcie v desiatich riadkov\u00fdch vet\u00e1ch s formulami, ktor\u00e9 s\u00fa prevzat\u00e9 'zo stropu' \u2013 to je t\u00e9ma na samostatn\u00fd \u010dl\u00e1nok (probl\u00e9m <noindex><a rel=\"nofollow\" href=\"https:\/\/en.wikipedia.org\/wiki\/Publish_or_perish\">publish or perish<\/a><\/noindex>).<\/p>\n<p><\/p>\n<p>Z tohto d\u00f4vodu vznikla potreba systematizova\u0165 inform\u00e1cie o neur\u00f3nov\u00fdch sie\u0165ach a t\u00fdm zv\u00fd\u0161i\u0165 kvalitu pochopenia a zapam\u00e4tania. Hlavnou t\u00e9mou anal\u00fdzy jednotliv\u00fdch technol\u00f3gi\u00ed a architekt\u00far umel\u00fdch neur\u00f3nov\u00fdch siet\u00ed sa stala t\u00e1to \u00faloha: <strong>zisti\u0165, kam to v\u0161etko smeruje<\/strong>, a nie usporiadanie nejakej konkr\u00e9tnej neur\u00f3novej siete samostatne.<\/p>\n<p><\/p>\n<p>Kam v\u0161etko to smeruje. Hlavn\u00e9 v\u00fdsledky:<\/p>\n<p><\/p>\n<ul>\n<li>Po\u010det startupov v oblasti strojov\u00e9ho u\u010denia v posledn\u00fdch dvoch rokoch <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/recognitor\/blog\/455676\/\">v\u00fdrazne klesol<\/a><\/noindex>. Mo\u017en\u00e1 pr\u00ed\u010dina: 'neur\u00f3nov\u00e9 siete prestali by\u0165 nie\u010d\u00edm nov\u00fdm'.<\/li>\n<li>Ka\u017ed\u00fd bude schopn\u00fd vytvori\u0165 funguj\u00facu neur\u00f3nov\u00fa sie\u0165 na rie\u0161enie jednoduch\u00e9ho probl\u00e9mu. Na to si vezme hotov\u00fd model z 'zoo modelov' (model zoo) a natr\u00e9nuje posledn\u00fa vrstvu neur\u00f3novej siete (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/yofjFQddwHE\">transfer learning<\/a><\/noindex>) na hotov\u00fdch d\u00e1tach z <noindex><a rel=\"nofollow\" href=\"https:\/\/toolbox.google.com\/datasetsearch\">Google Dataset Search<\/a><\/noindex> alebo z <noindex><a rel=\"nofollow\" href=\"https:\/\/www.kaggle.com\/datasets\">25-tis\u00edc datasetov Kaggle<\/a><\/noindex> v bezplatnom <noindex><a rel=\"nofollow\" href=\"https:\/\/www.dataschool.io\/cloud-services-for-jupyter-notebook\/\">obla\u010dnom prostred\u00ed Jupyter Notebook<\/a><\/noindex>.<\/li>\n<li>Ve\u013ek\u00ed v\u00fdrobcovia neur\u00f3nov\u00fdch siet\u00ed za\u010dali vytv\u00e1ra\u0165 <strong>'zoo modelov'<\/strong> (model zoo). S ich pomocou je mo\u017en\u00e9 r\u00fdchlo vytvori\u0165 komer\u010dn\u00fa aplik\u00e1ciu: <noindex><a rel=\"nofollow\" href=\"https:\/\/tfhub.dev\/\">TF Hub<\/a><\/noindex> pre TensorFlow, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/open-mmlab\/mmdetection\">MMDetection<\/a><\/noindex> pre PyTorch, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/facebookresearch\/Detectron\">Detectron<\/a><\/noindex> pre Caffe2, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/wkentaro\/chainer-modelzoo\">chainer-modelzoo<\/a><\/noindex> pre Chainer a <noindex><a rel=\"nofollow\" href=\"https:\/\/modelzoo.co\/\">in\u00e9<\/a><\/noindex>.<\/li>\n<li>Neur\u00f3nov\u00e9 siete funguj\u00face v <strong>re\u00e1lnom \u010dase<\/strong> na mobiln\u00fdch zariadeniach. Od 10 do 50 sn\u00edmok za sekundu.<\/li>\n<li>Pou\u017eitie neur\u00f3nov\u00fdch siet\u00ed v telef\u00f3noch (TF Lite), v prehliada\u010doch (TF.js) a v <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/19ZNz2N79u4\">dom\u00e1cich predmetoch<\/a><\/noindex> (IoT, <strong>I<\/strong>internet <strong>o<\/strong>f <strong>T<\/strong>vec\u00ed). Obzvl\u00e1\u0161\u0165 v telef\u00f3noch, ktor\u00e9 u\u017e podporuj\u00fa neur\u00f3nov\u00e9 siete na \u00farovni 'hardv\u00e9ru' (neur\u00f3ny akceler\u00e1tory).<\/li>\n<li>Ka\u017ed\u00e9 zariadenie, oble\u010denie a mo\u017eno aj jedlo bud\u00fa ma\u0165 <strong>IP-v6 adresu<\/strong> a bud\u00fa si medzi sebou komunikova\u0165\" - <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/GG7H8Xa4m8I?t=85\">Sebasti\u00e1n Thrun<\/a><\/noindex>.<\/li>\n<li>Rast po\u010dtu publik\u00e1ci\u00ed o strojovom u\u010den\u00ed za\u010dal <noindex><a rel=\"nofollow\" href=\"http:\/\/data-mining.philippe-fournier-viger.com\/too-many-machine-learning-papers\">presahova\u0165 z\u00e1kon Moora<\/a><\/noindex> (zdvojenie ka\u017ed\u00e9 dva roky) od roku 2015. Zjavne s\u00fa potrebn\u00e9 neur\u00f3nov\u00e9 siete na anal\u00fdzu \u010dl\u00e1nkov.<\/li>\n<li>Nasleduj\u00face technol\u00f3gie z\u00edskavaj\u00fa popularitu:\n<ul>\n<li><strong>PyTorch<\/strong> \u2013 popularita rastie r\u00fdchlo a zd\u00e1 sa, \u017ee predbehne TensorFlow.<\/li>\n<li>Automatick\u00e1 optimaliz\u00e1cia hyperparametrov <strong>AutoML<\/strong> \u2013 popularita rastie postupne.<\/li>\n<li>Postupn\u00e9 zni\u017eovanie presnosti a zvy\u0161ovanie r\u00fdchlosti v\u00fdpo\u010dtov: <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rln_kZbYaWc\">fuzzy logika<\/a><\/noindex>, algoritmy <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/MIPkK5ZAsms\">boostingu<\/a><\/noindex>, nepresn\u00e9 (pribli\u017en\u00e9) v\u00fdpo\u010dty, kvantiz\u00e1cia (ke\u010f s\u00fa v\u00e1hy neur\u00f3nov\u00fdch siet\u00ed prev\u00e1dzan\u00e9 na cel\u00e9 \u010d\u00edsla a kvantizovan\u00e9), neuroakceler\u00e1tory.<\/li>\n<li>Preklad <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">obr\u00e1zku do textu<\/a><\/noindex> a <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">textu do obr\u00e1zku<\/a><\/noindex>.<\/li>\n<li>Vytvorenie <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/OrHLacCDZVQ\">3D objektov z videa<\/a><\/noindex>, teraz u\u017e v re\u00e1lnom \u010dase.<\/li>\n<li>Hlavn\u00e9 v DL - to s\u00fa ve\u013ek\u00e9 d\u00e1ta, ale ich zber a ozna\u010dovanie nie je jednoduch\u00e9. Preto sa rozv\u00edja automatiz\u00e1cia ozna\u010dovania (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/NcKTn4C91Yc\">automated annotation<\/a><\/noindex>) pre neur\u00f3nov\u00e9 siete za pomoci neur\u00f3nov\u00fdch siet\u00ed.<\/li>\n<\/ul>\n<\/li>\n<li>S neur\u00f3nov\u00fdmi sie\u0165ami sa Computer Science n\u00e1hle stala <strong>experiment\u00e1lnou vedou<\/strong> a objavil sa <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/480348\">kr\u00edzov\u00fd stav reprodukovate\u013enosti.<\/a><\/noindex>.<\/li>\n<li>IT peniaze a popularita neur\u00f3nov\u00fdch siet\u00ed sa objavili z\u00e1rove\u0148, ke\u010f sa v\u00fdpo\u010dty stali trhovou hodnotou. Ekonomika zo zlatoko\u0161ov\u00e9ho typu prech\u00e1dza <strong>na zlato-peniazo-v\u00fdpo\u010dtov\u00fd typ.<\/strong>Pozrite sa na m\u00f4j \u010dl\u00e1nok o <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\">ekonofyzike<\/a><\/noindex> a d\u00f4vodoch vzniku IT pe\u0148az\u00ed.<\/li>\n<\/ul>\n<p><\/p>\n<p>Postupne sa objavuje nov\u00e1 <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/481844\">metodol\u00f3gia programovania ML\/DL<\/a><\/noindex> (Strojov\u00e9 u\u010denie &amp; Hlbok\u00e9 u\u010denie), ktor\u00e1 je zalo\u017een\u00e1 na predstaven\u00ed programu ako s\u00faboru tr\u00e9novan\u00fdch neur\u00f3nov\u00fdch modelov.<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neur\u00f3nov\u00e9 siete. Kam to v\u0161etko smeruje\" src=\"\/wp-content\/uploads\/2020\/01\/17df76a613191522145abeba177c2685.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Obr\u00e1zok 3 \u2013 ML\/DL ako nov\u00e1 metodol\u00f3gia programovania.<\/p>\n<p><\/p>\n<p>Av\u0161ak dodnes sa neobjavila <strong>\"te\u00f3ria neur\u00f3nov\u00fdch siet\u00ed\"<\/strong>, v r\u00e1mci ktorej je mo\u017en\u00e9 myslie\u0165 a pracova\u0165 syst\u00e9movo. To, \u010do sa dnes naz\u00fdva \"te\u00f3riou\", s\u00fa v skuto\u010dnosti experiment\u00e1lne, heuritick\u00e9 algoritmy.<\/p>\n<p><\/p>\n<p>Odkazy na moje a nielen moje zdroje:<\/p>\n<p><\/p>\n<ul>\n<li>Newsletter o Data Science. Hlavne o spracovan\u00ed obrazov. Kto chce dost\u00e1va\u0165, nech po\u0161le e-mail (foobar167&lt;gaff-gaff&gt;gmail&lt;dot&gt;com). Odkazy na \u010dl\u00e1nky a vide\u00e1 posielam pod\u013ea toho, ako sa hromadia materi\u00e1ly.<\/li>\n<li>Celkov\u00fd <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Machine_Learning\/courses_on_machine_learning.md\">zoznam kurzov a \u010dl\u00e1nkov<\/a><\/noindex>, ktor\u00e9 som na\u0161tudoval a ktor\u00e9 by som chcel absolvova\u0165.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Ubuntu\/13_Keras_and_TensorFlow_how-tos.md#exercises\">Kurzy a vide\u00e1 pre za\u010diato\u010dn\u00edkov<\/a><\/noindex>, s ktor\u00fdmi by sa malo za\u010da\u0165 pri u\u010den\u00ed neur\u00f3nov\u00fdch siet\u00ed. Plus bro\u017e\u00fara <noindex><a rel=\"nofollow\" href=\"https:\/\/foobar167.github.io\/page\/vvedeniye-v-mashinnoye-obucheniye-i-iskusstvennyye-neyronnyye-seti.html\">\"\u00davod do strojov\u00e9ho u\u010denia a umel\u00fdch neur\u00f3nov\u00fdch siet\u00ed\"<\/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\">U\u017eito\u010dn\u00e9 n\u00e1stroje<\/a><\/noindex>, kde si ka\u017ed\u00fd n\u00e1jde nie\u010do zauj\u00edmav\u00e9 pre seba.<\/li>\n<li>Extr\u00e9mne u\u017eito\u010dn\u00e9 sa uk\u00e1zali <strong>videokan\u00e1ly o anal\u00fdze vedeck\u00fdch \u010dl\u00e1nkov<\/strong> o Data Science. N\u00e1jdite ich, prihl\u00e1ste sa na odber a po\u0161lite odkazy svojim kolegom a mne tie\u017e. Pr\u00edklady:\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\">Blog Chengwei Zhang<\/a><\/noindex> ako <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/Tony607\">Tony607<\/a><\/noindex> s podrobn\u00fdmi pokynmi a otvoren\u00fdm zdrojov\u00fdm k\u00f3dom.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p><\/p>\n<p>\u010eakujem za pozornos\u0165!<\/p>\n<p>Zdroj: <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 5.0.4 - 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\/sk\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya\" \/>\n\t\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.4\" \/>\n\t\t<meta property=\"og:locale\" content=\"sk_SK\" \/>\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\/sk\/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\udd47Neur\u00f3nov\u00e9 siete. Kam to v\u0161etko smeruje | ProHoster","description":"\u010cl\u00e1nok sa sklad\u00e1 z dvoch \u010dast\u00ed: Stru\u010dn\u00fd preh\u013ead niektor\u00fdch architekt\u00far siet\u00ed na detekciu objektov na obr\u00e1zku a segment\u00e1ciu obr\u00e1zkov s najzrozumite\u013enej\u0161\u00edmi odkazmi na zdroje pre m\u0148a.","canonical_url":"https:\/\/prohoster.info\/sk\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"sk_SK","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\/sk\/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\/sk\/wp-json\/wp\/v2\/posts\/54780","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/comments?post=54780"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/posts\/54780\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/media?parent=54780"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/categories?post=54780"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/sk\/wp-json\/wp\/v2\/tags?post=54780"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}