Lapho ongaya khona: imicimbi ezayo yamahhala yochwepheshe be-IT eMoscow (Januwari 14-18)

Imicimbi enokubhaliswa okuvulekile:

• I-AI neselula
• I-NeurIPS Iphathi Yonyaka Omusha
• R Moscow Meetup #5
• I-Frontend Meetup e-Moscow (Gastromarket Balchug)
• Ukuqeqeshwa kokufunda ngomshini ku-Avito 2.0

I-AI neselula

Januwari 14, 19:00-22:00, ngoLwesibili

Sikumema emhlanganweni omayelana nobuhlakani bokwenziwa, ukusetshenziswa kwabo kumadivayisi eselula kanye nemikhuba ebaluleke kakhulu yezobuchwepheshe neyebhizinisi yeshumi leminyaka elisha. Uhlelo luhlanganisa imibiko ethokozisayo, izingxoxo, i-pizza kanye nomoya omuhle.

Esinye sezikhulumi siyingqalabutho ekwethuleni ubuchwepheshe bakamuva eHollywood, i-White House; incwadi yakhe ethi “Augmented: Life in the Smart Lane” kukhulunywe ngayo njengenye yezincwadi zakhe azikhonzile uMongameli waseChina enkulumweni yakhe yonyaka omusha.

I-NeurIPS Iphathi Yonyaka Omusha

Januwari 15, kusukela ngo-18:00, ngoLwesithathu

• 18:00 Ukubhalisa
• 19:00 Ukuvulwa - Mikhail Bilenko, Yandex
• 19:05 Ukufunda okuqinisiwe kwa-NeurIPS 2019: bekunjani - Sergey Kolesnikov, TinkoffMinyaka yonke isihloko se-reinforcement learning (RL) siya sishisa futhi sishube kakhulu. Futhi minyaka yonke, i-DeepMind ne-OpenAI zengeza uphethiloli emlilweni ngokukhulula i-bot entsha yokusebenza okungaphezu komuntu. Ingabe kukhona okuzuzisa ngempela ngemva kwalokhu? Futhi yimaphi amathrendi akamuva kukho konke ukuhlukahluka kwe-RL? Ake sithole!
• 19:25 Ukubuyekezwa komsebenzi we-NLP kwa-NeurIPS 2019 - Mikhail Burtsev, MIPTNamuhla, izitayela zokuphumelela kakhulu emkhakheni wokucubungula ulimi lwemvelo zihlotshaniswa nokwakhiwa kwezakhiwo ezisekelwe kumamodeli olimi namagrafu olwazi. Umbiko uzohlinzeka ngemininingwane yemisebenzi lapho lezi zindlela zisetshenziswa khona ukwakha amasistimu ezingxoxo ukuze kuqaliswe imisebenzi eyahlukene. Isibonelo, ukuxhumana ngezihloko ezijwayelekile, ukukhulisa ukuzwelana nokuqhuba inkhulumomphendvulwano egxile emgomweni.
• 19:45 Izindlela zokuqonda uhlobo lomsebenzi wokulahlekelwa - Dmitry Vetrov, Faculty of Computer Science, National Research University Higher School of EconomicsNgizoxoxa ngamaphepha amaningana ahlola imiphumela engajwayelekile ekufundeni okujulile. Le miphumela isikhanyisela ekubukeni kobuso bomsebenzi wokulahlekelwa endaweni yesisindo futhi isivumele ukuba sibeke phambili inani lemibono. Uma kuqinisekisiwe, kuzokwazi ukulawula ngempumelelo usayizi wesinyathelo ezindleleni zokuthuthukisa. Lokhu kuzokwenza kube nokwenzeka ukubikezela inani elifinyelelekayo lomsebenzi wokulahlekelwa kusampula yokuhlola isikhathi eside ngaphambi kokuphela kokuqeqeshwa.
• 20:05 Ukubuyekezwa kwemisebenzi yombono wekhompyutha kwa-NeurIPS 2019 - Sergey Ovcharenko, Konstantin Lakhman, YandexSizobheka izindawo eziyinhloko zocwaningo kanye nokusebenza ngombono wekhompyutha. Ake sizame ukuqonda ukuthi ngabe zonke izinkinga sezixazululiwe kusukela ekubukeni kwe-academy, noma ngabe imashi yokunqoba ye-GAN iyaqhubeka kuzo zonke izindawo, ngubani omelene nayo, futhi lapho ukuguqulwa okungagadiwe kuzokwenzeka.
• 20:25 Ikhefu lekhofi
• 20:40 Ukumodela okulandelanayo ngokuhleleka okungenamkhawulo kwesizukulwane - Dmitry Emelianenko, YandexSiphakamisa imodeli engafaka amagama ezindaweni ezingafanele emshweni okhiqiziwe. Imodeli ifunda ngokungaguquki ukuhleleka kokukhishwa kwamakhodi okusekelwe kudatha. Ikhwalithi engcono kakhulu ifinyelelwa kumadathasethi ambalwa: ngokuhumusha komshini, sebenzisa ku-LaTeX nencazelo yesithombe. Umbiko unikezelwe esihlokweni lapho sibonisa khona ukuthi uhlelo lokuqopha olufundiwe lunengqondo futhi luqondene ngqo nenkinga exazululwayo.
• 20:55 Ukuqeqeshwa Okuhlehlisiwe kwe-KL-Divergence yamanethiwekhi Angaphambili: Ukungaqiniseki Okuthuthukisiwe kanye Nokuqina Okuqinile - Andrey Malinin, YandexIzindlela zokuhlanganisa zokulinganisa ukungaqiniseki zisanda kusetshenziswa emisebenzini yokutholwa kwezigaba ezingezinhle, ukuthola okokufaka okungaphandle kokusabalalisa kanye nokutholwa kokuhlaselwa kwezitha. Amanethiwekhi Angaphambili aphakanyiswe njengendlela yokulingisa ngempumelelo iqoqo lamamodeli ukuze ahlukaniswe ngokubeka ipharamitha ye-Dirichlet ngaphambi kokusatshalaliswa kokusatshalaliswa kokuphumayo. Lawa mamodeli aboniswe ukuthi asebenza kahle kakhulu kunezinye izindlela zokuhlanganisa, ezifana ne-Monte-Carlo Dropout, emsebenzini wokuthola okokufaka ngaphandle kokusatshalaliswa. Nokho, ukukala Amanethiwekhi Angaphambili kumadathasethi ayinkimbinkimbi anamakilasi amaningi kunzima usebenzisa indlela yokuqeqesha eyayihlongozwe ekuqaleni. Leli phepha lenza iminikelo emibili. Okokuqala, sibonisa ukuthi umbandela wokuqeqeshwa ofanelekile wamaNethiwekhi Angaphambili ukuhlehla kwe-KL-divergence phakathi kokusabalalisa kwe-Dirichlet. Lezi zinkinga zidingida uhlobo lokusatshalaliswa kwethagethi yedatha yokuqeqeshwa, okuvumela amanethiwekhi angaphambili ukuthi aqeqeshwe ngempumelelo emisebenzini yokuhlukanisa ngezigaba eziningi ngokungafanele, kanye nokuthuthukisa ukusebenza kokutholwa ngaphandle kokusatshalaliswa. Okwesibili, sisebenzisa lesi simiso esisha sokuqeqesha, leli phepha liphenya kusetshenziswa amaNethiwekhi Angaphambili ukuze kutholwe ukuhlaselwa kwezitha futhi liphakamise uhlobo olujwayelekile lokuqeqeshwa kwezitha. Kuboniswa ukuthi ukwakhiwa kokuhlaselwa kwebhokisi elimhlophe okusebenzayo okuphumelelayo, okuthinta ukubikezela nokugwema ukutholwa, ngokumelene namaNethiwekhi Angaphambili aqeqeshwe ku-CIFAR-10 kanye ne-CIFAR-100 kusetshenziswa indlela ehlongozwayo kudinga inani elikhulu lomzamo wokuhlanganisa kunamanethiwekhi avikelwe kusetshenziswa izitha ezijwayelekile. ukuqeqeshwa noma ukuyeka i-MC.
• 21:10 Ingxoxo yephaneli: “NeurlPS, eye yanda kakhulu: ubani okufanele asolwe futhi enzeni?” - Alexander Krainov, Yandex
• 21:40 Iphathi

R Moscow Meetup #5

Januwari 16, 18:30-21:30, ngoLwesine

• 19:00-19:30 "Ukuxazulula izinkinga zokusebenza kusetshenziswa i-R yama-dummies" - u-Konstantin Firsov (Netris JSC, Unjiniyela Omkhulu Wokusebenza).
• 19: 30-20: 00 "Ukuthuthukiswa kwe-inventory ezitolo" - u-Genrikh Ananyev (Iqembu le-PJSC Beluga, Inhloko yokubika okuzenzakalelayo).
• 20:00-20:30 “I-BMS ku-X5: kwenziwa kanjani izimayini zenqubo yebhizinisi ezingodweni ze-POS ezingahlelekile kusetshenziswa i-R” - U-Evgeniy Roldugin (I-X5 Retail Group, Inhloko Yomnyango Wamathuluzi Wokulawula Ikhwalithi Yesevisi), u-Ilya Shutov (I-Media Tel, Inhloko wesazi sedatha yoMnyango).

I-Frontend Meetup e-Moscow (Gastromarket Balchug)

NgoJanuwari 18, 12:00-18:00, ngoMgqibelo

• "Kufanele nini ukubhala kabusha isicelo kusukela ekuqaleni, futhi kanjani ukukholisa ibhizinisi ngalokhu" - Alexey Pyzhyanov, umthuthukisi, SiburIndaba yangempela yokuthi sibhekane kanjani nezikweletu zobuchwepheshe ngendlela eqine kakhulu. Ngizokutshela ngakho:
  1. Kungani isicelo esihle siphenduke ifa elibi.
  2. Indlela esenze ngayo isinqumo esinzima sokubhala kabusha yonke into.
  3. Siwuthengise kanjani lo mbono kumnikazi womkhiqizo.
  4. Yini eyaphuma kulo mbono ekugcineni, nokuthi kungani singazisoli ngesinqumo esasenza.

• "I-Vuejs API iyahlekisa" - Vladislav Prusov, unjiniyela we-Frontind, AGIMA

Ukuqeqeshwa kokufunda ngomshini ku-Avito 2.0

NgoJanuwari 18, 12:00-15:00, ngoMgqibelo

• 12:00 “Zindi Sendy Logistics Challenge (rus)” - Roman Pyankov
• 12:30 “Data Souls Wildfire AI (rus)” - U-Ilya Plotnikov
• 13:00 Ikhefu lekhofi
• 13:20 “Topcoder SpaceNet 5 Challenge & Signate The 3rd Tellus Satellite Challenge (eng)” - Ilya Kibardin
• 14:00 Ikhefu lekhofi
• 14:10 “Codalab Automated Time Series Regression (eng)” — uDenis Vorotyntsev

Source: www.habr.com