{"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\/et\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","title":{"rendered":"Tehisintellekt. Kuhu see k\u00f5ik liigub","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>Artikkel koosneb kahest osast:<\/p>\n<p><\/p>\n<ol>\n<li>L\u00fchike \u00fclevaade m\u00f5nest objektide tuvastamise ja pildisegmentatsiooni v\u00f5rgustiku arhitektuurist koos k\u00f5ige arusaadavamate linkidega ressurssidele. P\u00fc\u00fcdsin valida videote selgitusi ja eelistatavalt vene keeles.<\/li>\n<li>Teine osa seisneb katses m\u00f5ista n\u00e4rviv\u00f5rkude arhitektuuride arengusuunda. Ja nende p\u00f5hjal arenevate tehnoloogiate.<\/li>\n<\/ol>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Tehisintellekt. Kuhu see k\u00f5ik liigub\" src=\"\/wp-content\/uploads\/2020\/01\/3e0238547dc956dbb069c11241e4534f.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Joonis 1 \u2013 N\u00e4rviv\u00f5rkude arhitektuuride m\u00f5istmine ei ole lihtne<\/p>\n<p><\/p>\n<p>K\u00f5ik algas sellest, et tegin kaks demo rakendust objekti klassifitseerimiseks ja tuvastamiseks Android telefonis:<\/p>\n<p><\/p>\n<ul>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/junkyard\/tree\/master\/object_classifier\">Tagaplaanidemo<\/a><\/noindex>, kus andmed t\u00f6\u00f6tlevad serveris ja edastatakse telefonile. Piltide klassifitseerimine (image classification) kolme t\u00fc\u00fcpi karupoegade puhul: pruuni, musta ja pl\u00fc\u00fcsiga.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/android\/tree\/master\/object_detection_demo\">Esiplaanidemo<\/a><\/noindex>, kus andmed t\u00f6\u00f6tlevad telefonis. Objektide tuvastamine (object detection) kolme t\u00fc\u00fcpi: sarapuu, viigimarja ja date.<\/li>\n<\/ul>\n<p><noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex><\/p>\n<p>Klassifitseerimise, objekti tuvastamise ja <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/analytics-vidhya\/image-classification-vs-object-detection-vs-image-segmentation-f36db85fe81\">pildisegmentatsiooni<\/a><\/noindex>\u00fclesannete vahel on erinevus. Seet\u00f5ttu tekkis vajadus teada, millised n\u00e4rviv\u00f5rgu arhitektuurid tuvastavad objekte piltidel ja millised v\u00f5ivad segmenteerida. Leidsin j\u00e4rgmised arhitektuuride n\u00e4idised koos k\u00f5ige arusaadavamate linkidega ressurssidele:<\/p>\n<p><\/p>\n<ul>\n<li>R-CNN (<strong>R<\/strong>regioonid koos <strong>C<\/strong>konvolutsiooni <strong>N<\/strong>neuraalv\u00f5rkude omadustega): R-CNN, Fast R-CNN, <strong>N<\/strong>Faster R-CNN <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/@smallfishbigsea\/faster-r-cnn-explained-864d4fb7e3f8\">Mask R-CNN<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/0vt05rQqk_I\">. Objektide tuvastamiseks pildil kasutatakse piirkonna ettepaneku v\u00f5rku (Region Proposal Network, RPN), mis eristab piiratud piirkondi (bounding boxes). Alguses kasutati RPN asemel aeglasemat mehhanismi Selective Search. Seej\u00e4rel antakse eristatud piiratud piirkonnad tavaliseks n\u00e4rviv\u00f5rguks klassifitseerimise jaoks. R-CNN arhitektuuris on selged \u00abfor\u00bb ts\u00fcklid eristatud piirkondade l\u00e4bit\u00f6\u00f6tamiseks, k\u00f5igi 2000 sisendi l\u00e4bimise kaudu sisemise AlexNet v\u00f5rgustiku kaudu. Selgete \u00abfor\u00bb ts\u00fcklite t\u00f5ttu aeglustub piltide t\u00f6\u00f6tlemise kiirus. Selgete ts\u00fcklite arv, sisemise n\u00e4rviv\u00f5rgu l\u00e4bimised, v\u00e4heneb koos iga uue arhitektuuri versiooniga, samuti tehakse k\u00fcmneid muid muudatusi kiirusel ja objektide tuvastamise \u00fclesande asendamisel objektide segmentimisega Mask R-CNN-is.<\/a><\/noindex>YOLO<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/L0tzmv--CGY\">ainult<\/a><\/noindex> (<strong>Y<\/strong>vaatama <strong>O<\/strong>ainult <strong>L<\/strong>vaatama <strong>O<\/strong>YOLO (You Only Look Once) \u2013 esimene n\u00e4rviv\u00f5rk, mis tuvastas objekte reaalajas mobiilseadmetes. Erip\u00e4ra: objektide eristamine \u00fche korra jooksul (piisab, kui vaadata \u00fcks kord). See t\u00e4hendab, et YOLO arhitektuuris ei ole selgeid \u201efor\u201d silmuseid, mist\u00f5ttu v\u00f5rk t\u00f6\u00f6tab kiiresti. N\u00e4iteks sarnaneb see NumPy-ga, kus maatriksite toimingutes ei ole selgeid \u201efor\u201d silmuseid, mis nende madalamal tasemel C programmeerimiskeeles teostatakse. YOLO kasutab eeldefineeritud akende v\u00f5rgustikku. Korraga sama objekti tuvastamiseks kasutatakse akende kattuvuse koefitsienti (IoU, <strong>I<\/strong>ristumist <strong>o<\/strong>\u00fcle <strong>U<\/strong>liitu). Antud arhitektuur t\u00f6\u00f6tab laias valikus ja omab k\u00f5rget <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\">robustsust<\/a><\/noindex>: mudel saab olla koolitatud fotode peal, kuid t\u00f6\u00f6tab h\u00e4sti ka joonistatud piltidel.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/P8e-G-Mhx4k\">SSD<\/a><\/noindex> (<strong>S<\/strong>Single <strong>S<\/strong>Shot MultiBox <strong>D<\/strong>Detector) \u2013 kasutatakse k\u00f5ige \u00f5nnestunumaid \u201enippe\u201d YOLO arhitektuurist (n\u00e4iteks non-maximum suppression) ja lisatakse uusi, et n\u00e4rviv\u00f5rk t\u00f6\u00f6taks kiiremini ja t\u00e4psemalt. Erip\u00e4ra: objektide eristamine \u00fche korra jooksul antud akende v\u00f5rgustiku (default box) abil pildip\u00fcramiidis. Pildip\u00fcramiid on kodeeritud konvolutsiooniliste tensoreid j\u00e4rjestikuste konvolutsioonide ja maksimeerimise toimingute kaudu (maksimeerimise toimingute k\u00e4igus ruumiline dimensioon v\u00e4heneb). Nii tuvastatakse nii suuri kui ka v\u00e4ikseid objekte \u00fche korra jooksul.<\/li>\n<li>MobileSSD (<strong>Mobiilne<\/strong>NetV2 + <strong>SSD<\/strong>) \u2013 kahe n\u00e4rviv\u00f5rgu arhitektuuri kombinatsioon. Esimene v\u00f5rk <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/352804\/\">MobileNetV2<\/a><\/noindex> t\u00f6\u00f6tab kiiresti ja suurendab tuvastamise t\u00e4psust. MobileNetV2 kasutatakse VGG-16 asemel, mis algselt kasutati <noindex><a rel=\"nofollow\" href=\"https:\/\/arxiv.org\/abs\/1512.02325\">originaalartiklis<\/a><\/noindex>. Teine v\u00f5rk SSD tuvastab objektide asukoha pildil.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/ge_RT5wvHvY\">SqueezeNet<\/a><\/noindex> \u2013 v\u00e4ga v\u00e4ike, kuid t\u00e4pne n\u00e4rviv\u00f5rk. See ei lahenda iseseisvalt objektide tuvastamise \u00fclesannet. Kuid seda saab kasutada erinevate arhitektuuride kombinatsioonis. Ja kasutada mobiilseadmetes. Erip\u00e4raks on asjaolu, et k\u00f5igepealt andmed kokku surutakse nelja 1\u00d71 konvolutsioonifiltri abil ning seej\u00e4rel laienevad nelja 1\u00d71 ja nelja 3\u00d73 konvolutsioonifiltri abil. \u00dchte sellist andmete kokku surumise ja laienemise iteratsiooni nimetatakse \u201eFire Module\u201d.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/b6jhopSMit8\">DeepLab<\/a><\/noindex> (Semantic Image Segmentation with Deep Convolutional Nets) \u2013 objektide segmenteerimine pildil. Arhitektuuri iseloomulikuks tunnuseks on h\u00f5re konvolutsioon (dilated convolution), mis s\u00e4ilitab ruumilise eraldusv\u00f5ime. Seej\u00e4rel j\u00e4rgneb tulemuste j\u00e4relt\u00f6\u00f6tlus graafilise t\u00f5en\u00e4osusmudeli (conditional random field) abil, mis v\u00f5imaldab eemaldada v\u00e4ikseid m\u00fcrasid segmentatsioonist ja parandada segmentitud pildi kvaliteeti. Hirmu\u00e4ratava nimega \u201egraafiline t\u00f5en\u00e4osusmudel\u201d peidab end tavaline Gaussi filter, mis on ligikaudne viie punkti j\u00e4rgi.<\/li>\n<li>P\u00fc\u00fcdsin aru saada seadmest. <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), kuid ei m\u00f5istnud palju.<\/li>\n<li>Vaatasin ka, kuidas t\u00f6\u00f6tab \u201et\u00e4helepanu\u201d (attention) tehnoloogia: <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>. Arhitektuuri \u201et\u00e4helepanu\u201d iseloomulik tunnus on k\u00f5rgendatud t\u00e4helepanu piirkondade automaatne esiletoomine pildil (RoI, <strong>R<\/strong>egions <strong>o<\/strong>f <strong>I<\/strong>nterest) n\u00e4rviv\u00f5rgu Attention Unit abil. K\u00f5rgendatud t\u00e4helepanu piirkonnad sarnanevad piiratud piirkondadele (bounding boxes), kuid erinevalt neist ei ole nad pildil fikseeritud ja v\u00f5ivad omada uduseid piire. Seej\u00e4rel eraldatakse k\u00f5rgendatud t\u00e4helepanu piirkondadest omadused (feature), mis \u201etoidetakse\u201d rekursiivsetele n\u00e4rviv\u00f5rkudele arhitektuuridega <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/5lUUrREboSk\">LSDM, GRU v\u00f5i Vanilla RNN<\/a><\/noindex>. Rekursiivsed n\u00e4rviv\u00f5rgud oskavad anal\u00fc\u00fcsida omaduste suhteid j\u00e4rjestuses. Rekursiivseid n\u00e4rviv\u00f5rke kasutati algselt teksti t\u00f5lkimiseks teistesse keeltesse, n\u00fc\u00fcd aga ka <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">pildi t\u00f5lkimiseks tekstiks,<\/a><\/noindex> ja <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">teksti t\u00f5lkimiseks pildiks.<\/a><\/noindex>.<\/li>\n<\/ul>\n<p><\/p>\n<p>Nende arhitektuuridega tutvudes <strong>aru, et ma ei m\u00f5ista midagi.<\/strong>. See ei ole tingitud sellest, et minu n\u00e4rviv\u00f5rgul oleks probleeme t\u00e4helepanu mehhanismiga. K\u00f5ikide nende arhitektuuride loomine sarnaneb m\u00f5ne tohutu hackathoniga, kus autorid konkureerivad h\u00e4kkimise oskustes. H\u00e4kk (hack) on kiire lahendus keerulisele programmik\u00fcsimusele. Ehkki nende k\u00f5igi arhitektuuride vahel puudub n\u00e4htav ja arusaadav loogiline seos. K\u00f5ik, mis neid \u00fchendab, on k\u00f5ige edukamate h\u00e4kete kogum, mida nad \u00fcksteiselt laenavad, pluss k\u00f5igile \u00fchine <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/Ilg3gGewQ5U\">tagasiside konvolutsiooni operatsioon<\/a><\/noindex> (tagasiside levitamine, backpropagation). Ei ole <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/272473\/\">s\u00fcsteemset m\u00f5tlemist!<\/a><\/noindex>Pole selge, mida muuta ja kuidas olemasolevaid saavutusi optimeerida.<\/p>\n<p><\/p>\n<p>Tulemuseks on loogilise seose puudumine haakide vahel, mist\u00f5ttu neid on \u00e4\u00e4rmiselt raske meeles pidada ja praktikas rakendada. Need on killustatud teadmised. Parimal juhul j\u00e4\u00e4b meelde paar huvitavat ja ootamatut hetke, kuid enamus m\u00f5istetust ja arusaamatust kaob m\u00e4lu s\u00fcgavustest juba m\u00f5ne p\u00e4eva p\u00e4rast. Olgu hea, kui n\u00e4dal p\u00e4rast j\u00e4\u00e4b meelde v\u00e4hemalt arhitektuuri nimi. Ja lugemiseks artiklite ja \u00fclevaatevideote vaatamiseks on kulutatud mitu tundi ja isegi p\u00e4evade kaupa t\u00f6\u00f6aega!<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Tehisintellekt. Kuhu see k\u00f5ik liigub\" src=\"\/wp-content\/uploads\/2020\/01\/86f0f24e3be1d0f1a210f3132897d981.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Joonis 2 \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/www.asimovinstitute.org\/neural-network-zoo\/\">Neuronaalsete v\u00f5rkude loomaaed<\/a><\/noindex><\/p>\n<p><\/p>\n<p>Enamik teadusartiklite autoreid, minu arvates, teeb k\u00f5ik, et isegi need killustatud teadmised j\u00e4\u00e4ksid lugejale arusaamatuks. Kuid gerundivormid k\u00fcmnes rida lauses, milles on valemid, mis on \u201elaest v\u00f5etud\u201d \u2013 see on teema eraldi artikliks (probleem <noindex><a rel=\"nofollow\" href=\"https:\/\/en.wikipedia.org\/wiki\/Publish_or_perish\">publish or perish<\/a><\/noindex>).<\/p>\n<p><\/p>\n<p>Selle t\u00f5ttu tekkis vajadus s\u00fcstematiseerida teave neuronaalsete v\u00f5rkude kohta ja seega suurendada arusaamise ja meeldej\u00e4tmise kvaliteeti. Seet\u00f5ttu sai p\u00f5hiliseks teema teatud tehnoloogiate ja tehisneuronaalsete v\u00f5rkude arhitektuuride anal\u00fc\u00fcsiks j\u00e4rgmine \u00fclesanne: <strong>teada, kuhu see k\u00f5ik liigub<\/strong>, mitte \u00fcksiku neuronaalv\u00f5rgu seadistust eraldi.<\/p>\n<p><\/p>\n<p>Kuhu see k\u00f5ik liigub. Peamised tulemused:<\/p>\n<p><\/p>\n<ul>\n<li>Masin\u00f5ppe valdkonnas alustanud startupide arv on viimase kahe aasta jooksul <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/recognitor\/blog\/455676\/\">m\u00e4rgatavalt v\u00e4henenud<\/a><\/noindex>. V\u00f5imalik p\u00f5hjus: \u201eneuronaalsed v\u00f5rgud pole enam midagi uut.\u201d<\/li>\n<li>Iga\u00fcks suudab luua t\u00f6\u00f6tava neuronaalv\u00f5rgu lihtsa \u00fclesande lahendamiseks. Selle jaoks v\u00f5etakse valmis mudel \"mudelite loomaaia\" (model zoo) seest ja treenitakse neuronaalv\u00f5rgu viimane kiht (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/yofjFQddwHE\">transfer learning<\/a><\/noindex>) valmis andmete p\u00f5hjal <noindex><a rel=\"nofollow\" href=\"https:\/\/toolbox.google.com\/datasetsearch\">Google Dataset Search<\/a><\/noindex> v\u00f5i <noindex><a rel=\"nofollow\" href=\"https:\/\/www.kaggle.com\/datasets\">25 000 Kaggle andmestikku<\/a><\/noindex> tasuta <noindex><a rel=\"nofollow\" href=\"https:\/\/www.dataschool.io\/cloud-services-for-jupyter-notebook\/\">Jupyter Notebooks<\/a><\/noindex>.<\/li>\n<li>Suured neuronaalv\u00f5rkude tootjad on hakanud looma <strong>\"mudelite loomaaedu\"<\/strong> (model zoo). Nende abil on v\u00f5imalik kiiresti luua kommertsrakendus: <noindex><a rel=\"nofollow\" href=\"https:\/\/tfhub.dev\/\">TF Hub<\/a><\/noindex> TensorFlow jaoks, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/open-mmlab\/mmdetection\">MMDetection<\/a><\/noindex> PyTorch jaoks, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/facebookresearch\/Detectron\">Detectron<\/a><\/noindex> Caffe2 jaoks, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/wkentaro\/chainer-modelzoo\">chainer-modelzoo<\/a><\/noindex> Chaineri jaoks ja <noindex><a rel=\"nofollow\" href=\"https:\/\/modelzoo.co\/\">muud<\/a><\/noindex>.<\/li>\n<li>Neuronaalv\u00f5rgud, mis t\u00f6\u00f6tavad <strong>reaalajas<\/strong> (real-time) mobiilseadmetes. Alates 10 kuni 50 kaadrit sekundis.<\/li>\n<li>Neuronaalv\u00f5rkude rakendamine telefonides (TF Lite), brauserites (TF.js) ja <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/19ZNz2N79u4\">koduvidinates<\/a><\/noindex> (IoT, <strong>I<\/strong>internet <strong>o<\/strong>f <strong>T<\/strong>asjad). Eriti telefonides, mis juba toetavad neuronaalv\u00f5rke riistvara tasemel (neuroakseleraatorid).<\/li>\n<li>Iga seade, riided ja v\u00f5ib-olla isegi toit omavad <strong>IP-v6 aadressi<\/strong> ja suhtlevad omavahel\" \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/GG7H8Xa4m8I?t=85\">Sebastian Thrun<\/a><\/noindex>.<\/li>\n<li>Masin\u00f5ppe publikatsioonide arvu kasv on hakanud <noindex><a rel=\"nofollow\" href=\"http:\/\/data-mining.philippe-fournier-viger.com\/too-many-machine-learning-papers\">\u00fcletama Moore'i seadust<\/a><\/noindex> (kordumine iga kahe aasta tagant) alates 2015. aastast. Ilmselt on vajalikud n\u00e4rviv\u00f5rgud artiklite anal\u00fc\u00fcsimiseks.<\/li>\n<li>Muganduvad populaarsust j\u00e4rgmised tehnoloogiad:\n<ul>\n<li><strong>PyTorch<\/strong> \u2013 populaarsus kasvab kiirelt ja n\u00e4ib, et \u00fcletab TensorFlow'd.<\/li>\n<li>Automaatne h\u00fcperparameetrite valik <strong>AutoML<\/strong> \u2013 populaarsus kasvab \u00fchtlaselt.<\/li>\n<li>T\u00e4psuse j\u00e4rkj\u00e4rguline v\u00e4henemine ja arvutuste kiirus: <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rln_kZbYaWc\">uudne loogika<\/a><\/noindex>, algoritmid <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/MIPkK5ZAsms\">boostimise<\/a><\/noindex>, ebat\u00e4psed (ligikaudsed) arvutused, kvantiseerimine (kui n\u00e4rviv\u00f5rkude kaalukad muudetakse t\u00e4isarvudeks ja kvantiseeritakse), neuroakseleraatorid.<\/li>\n<li>T\u00f5lge <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">pildi t\u00f5lkimiseks tekstiks,<\/a><\/noindex> ja <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">teksti t\u00f5lkimiseks pildiks.<\/a><\/noindex>.<\/li>\n<li>Meetodi loomine <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/OrHLacCDZVQ\">kolmem\u00f5\u00f5tmeliste objektide j\u00e4lgimine videolt<\/a><\/noindex>, n\u00fc\u00fcd juba reaalajas.<\/li>\n<li>DL-i p\u00f5hisisu on suur andmehulk, kuid nende kogumine ja m\u00e4rgistamine pole lihtne. Seet\u00f5ttu areneb automatiseeritud m\u00e4rgistamine (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/NcKTn4C91Yc\">automated annotation<\/a><\/noindex>) n\u00e4rviv\u00f5rkude jaoks, kasutades n\u00e4rviv\u00f5rke.<\/li>\n<\/ul>\n<\/li>\n<li>N\u00e4rviv\u00f5rkude puhul sai arvutiteadus \u00e4kitselt <strong>eksperimentaalseks teaduseks<\/strong> ja tekkis <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/480348\">reproduktiivsuse kriis.<\/a><\/noindex>.<\/li>\n<li>IT-raha ja n\u00e4rviv\u00f5rkude populaarsus ilmus korraga, kui arvutused muutusid turuhinnaga. Majandus kullavaluta muutub <strong>kulla-varude-arvutuseks.<\/strong>Vaata minu artiklit <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\">ekonofoorika<\/a><\/noindex> ja IT-raha tekkimise p\u00f5hjuse kohta.<\/li>\n<\/ul>\n<p><\/p>\n<p>Aeglaselt ilmub uus <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/481844\">ML\/DL programmeerimismetodoloogia<\/a><\/noindex> (Masin\u00f5pe &amp; S\u00fcva\u00f5pe), mis p\u00f5hineb programmi esitlemisel kui treenitud n\u00e4rviv\u00f5rkude mudelite kogumil.<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Tehisintellekt. Kuhu see k\u00f5ik liigub\" src=\"\/wp-content\/uploads\/2020\/01\/17df76a613191522145abeba177c2685.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>Joonis 3 \u2013 ML\/DL kui uus programmeerimismetodoloogia<\/p>\n<p><\/p>\n<p>Kuid siiski ei ole ilmunud <strong>\u00bbn\u00e4rviv\u00f5rkude teooriat\u00ab<\/strong>, mille p\u00f5hjal saaks m\u00f5elda ja s\u00fcsteemselt t\u00f6\u00f6tada. See, mida praegu nimetatakse \"teooriaks\", on tegelikult eksperimenteerimised, heuristilised algoritmid.<\/p>\n<p><\/p>\n<p>Viidatud minu ja teiste ressursside kohta:<\/p>\n<p><\/p>\n<ul>\n<li>Data Science uudiskiri. Peamiselt piltide t\u00f6\u00f6tlemise kohta. Kes soovib saada, saatke palun e-kiri (foobar167gmailcom). Artiklite ja videote lingid saadan kogunemise korral.<\/li>\n<li>\u00dchine <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Machine_Learning\/courses_on_machine_learning.md\">kursuste ja artiklite nimekiri<\/a><\/noindex>, mida olen l\u00e4binud ja mida sooviksin l\u00e4bida.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Ubuntu\/13_Keras_and_TensorFlow_how-tos.md#exercises\">Algajate kursused ja videod<\/a><\/noindex>, millega tasub alustada n\u00e4rviv\u00f5rkude \u00f5ppimist. Pluss bro\u0161\u00fc\u00fcr <noindex><a rel=\"nofollow\" href=\"https:\/\/foobar167.github.io\/page\/vvedeniye-v-mashinnoye-obucheniye-i-iskusstvennyye-neyronnyye-seti.html\">\u00bbSissejuhatus masin\u00f5ppesse ja kunstlikesse n\u00e4rviv\u00f5rkudesse\u00ab<\/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\">Kasulikud t\u00f6\u00f6riistad<\/a><\/noindex>, kus iga\u00fcks leiab midagi huvitavat enda jaoks.<\/li>\n<li>\u00c4\u00e4rmiselt kasulikud on olnud <strong>videokanaleid teadusartiklite anal\u00fc\u00fcsi kohta<\/strong> Data Science'i valdkonnas. Leidke neid, tellige ja jagage linke oma kolleegidega ning ka minuga. N\u00e4ited:\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 Zhangi blogi<\/a><\/noindex> ja <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/Tony607\">Tony607<\/a><\/noindex> samuti samm-sammult juhendite ja avatud koodiga.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p><\/p>\n<p>Ait\u00e4h t\u00e4helepanu eest!<\/p>\n<p>Allikas: <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 5.0.1.1 - 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\/et\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya\" \/>\n\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.1.1\" \/>\n\t\t<meta property=\"og:locale\" content=\"et_EE\" \/>\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\/et\/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\udd47N\u00e4rviv\u00f5rgud. Kuhu see k\u00f5ik suundub | ProHoster","description":"Artikkel koosneb kahest osast: L\u00fchike \u00fclevaade m\u00f5nest arhitektuurist objektide tuvastamiseks piltidel ja piltide segmentimiseks koos k\u00f5ige arusaadavamate linkidega ressurssidele.","canonical_url":"https:\/\/prohoster.info\/et\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"et_EE","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\/et\/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\/et\/wp-json\/wp\/v2\/posts\/54780","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/comments?post=54780"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/posts\/54780\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/media?parent=54780"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/categories?post=54780"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/et\/wp-json\/wp\/v2\/tags?post=54780"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}