Liketsahalo tse nang le ngoliso e bulehileng:
Pherekhong 14, 19:00-22:00, Labobeli
Re u memela sebokeng se mabapi le bohlale ba maiketsetso, ts'ebeliso ea bona ho lisebelisoa tsa mehala le mekhoa ea bohlokoahali ea thekenoloji le ea khoebo ea lilemo tse leshome tse ncha. Lenaneo le kenyelletsa litlaleho tse thahasellisang, lipuisano, pizza le maikutlo a monate.
E mong oa libui ke pula-maliboho ho hlahisa theknoloji ea morao-rao Hollywood, White House; buka ea hae "Augmented: Life in the Smart Lane" e ile ea boleloa e le e 'ngoe ea libuka tseo a li ratang haholo ke Mopresidente oa Chaena puong ea hae ea Selemo se Secha.
La 15 Pherekhong, ho qala ka 18:00, Laboraro
- 18:00 Ngoliso
- 19:00 Ho bula - Mikhail Bilenko, Yandex
- 19:05 Ho ithuta ho matlafatsa NeurIPS 2019: ho bile joang - Sergey Kolesnikov, TinkoffSelemo se seng le se seng sehlooho sa ho ithuta ka matla (RL) se ntse se chesa le ho feta. 'Me selemo se seng le se seng, DeepMind le OpenAI li eketsa mafura mollong ka ho lokolla bot e ncha ea ts'ebetso e phahametseng motho. Na ho na le ho hong ho hlileng ho molemo ka mor’a see? Hona ke mekhoa efe ea morao-rao ea mefuta eohle ea RL? A re boneng!
- 19:25 Tlhahlobo ea mosebetsi oa NLP ho NeurIPS 2019 - Mikhail Burtsev, MIPTKajeno, mekhoa e atlehang ka ho fetisisa lefapheng la ts'ebetso ea puo ea tlhaho e amahanngoa le kaho ea meralo e thehiloeng ho mehlala ea lipuo le li-graph tsa tsebo. Tlaleho e tla fana ka kakaretso ea mesebetsi eo mekhoa ena e sebelisoang ho aha mekhoa ea lipuisano ho phethahatsa mesebetsi e fapaneng. Ka mohlala, bakeng sa ho buisana ka lihlooho tse akaretsang, ho eketsa kutloelo-bohloko le ho etsa lipuisano tse nang le sepheo.
- 19:45 Litsela tsa ho utloisisa mofuta oa holim'a mosebetsi oa tahlehelo - Dmitry Vetrov, Faculty of Computer Science, National Research University Higher School of EconomicsKe tla tšohla lipampiri tse 'maloa tse hlahlobang liphello tse sa tloaelehang thutong e tebileng. Liphello tsena li fana ka leseli mabapi le ponahalo ea holim'a mosebetsi oa tahlehelo sebakeng sa boima ba 'mele' me li re lumella ho beha maikutlo a mangata. Haeba ho netefalitsoe, ho tla khonahala ho laola boholo ba mehato ka mokhoa o atlehileng haholoanyane ka mekhoa ea ho ntlafatsa. Sena se tla boela se etse hore ho khonehe ho bolela esale pele boleng bo ka finyelloang ba mosebetsi oa tahlehelo ho sampole ea teko nako e telele pele koetliso e fela.
- 20:05 Tlhahlobo ea mesebetsi ea pono ea khomphutha ho NeurIPS 2019 - Sergey Ovcharenko, Konstantin Lakhman, YandexRe tla sheba libaka tse ka sehloohong tsa lipatlisiso le ho sebetsa ponong ea k'homphieutha. A re leke ho utloisisa hore na mathata ohle a se a rarollotsoe ho ea ka pono ea sekolo, hore na leeto la tlhōlo la GAN le ntse le tsoela pele libakeng tsohle, ke mang ea hanyetsang, le hore na phetohelo e sa lebelloang e tla etsahala neng.
- 20:25 Kofi khefu
- 20:40 Ho etsa mohlala ka tatellano e se nang moeli ea moloko - Dmitry Emelianenko, YandexRe sisinya mohlala o ka kenyang mantsoe libakeng tse sa lumellaneng polelong e hlahisitsoeng. Moetso ona o ithuta ka mokhoa o hlakileng tatellano e bonolo ea decoding ho latela data. Boleng bo botle ka ho fetisisa bo fihlelleha ho li-dataset tse 'maloa: bakeng sa phetolelo ea mochine, sebelisa ho LaTeX le tlhaloso ea setšoantšo. Tlaleho e nehetsoe ho sengoloa seo ho sona re bonts'ang hore tatellano ea li-decoding e ithutoang e ea utloahala ebile e tobane le bothata bo rarolloang.
- 20:55 Reverse KL-Divergence Koetliso ea Li-Network tsa Pele: Ho Ntlafala ho Fetisisa le Matla a Bahanyetsi - Andrey Malinin, YandexMekhoa e kopaneng ea likhakanyo tse sa tsitsang e sa tsoa sebelisoa mesebetsing ea ho lemoha ho arohanngoa ka mokhoa o fosahetseng, ho lemoha lintho tse kentsoeng ka ntle ho kabo le ho lemoha tlhaselo ea bahanyetsi. Li-Network tsa Pele li khothalelitsoe e le mokhoa oa ho etsisa ka mokhoa o atlehileng sehlopha sa mefuta bakeng sa ho hlophisoa ka ho etsa parametering ea Dirichlet pele ho kabo ea lihlahisoa. Mefuta ena e bonts'itsoe hore e sebetsa hantle ho feta mekhoa e meng ea ho kopanya, joalo ka Monte-Carlo Dropout, mosebetsing oa ho lemoha lintho tse kentsoeng ka ntle ho kabo. Leha ho le joalo, ho lekanya Li-Network tsa Pele ho li-dataset tse rarahaneng tse nang le lihlopha tse ngata ho thata ho sebelisa mekhoa ea koetliso e neng e reriloe qalong. Letlapa lena le fana ka likarolo tse peli. Taba ea pele, re bonts'a hore mokhoa o nepahetseng oa koetliso bakeng sa Li-Network tsa Pele ke phapano ea KL lipakeng tsa kabo ea Dirichlet. Taba ena e bua ka mofuta oa kabo ea liphokotso tsa data ea lithupelo, e nolofalletsang marang-rang a pele ho koetlisoa ka katleho mesebetsing ea lihlopha ka lihlopha tse ngata, hammoho le ho ntlafatsa ts'ebetso ea ho lemoha ha e se e felile. Ea bobeli, ho nka monyetla ka mokhoa ona o mocha oa koetliso, pampiri ena e etsa lipatlisiso ka ho sebelisa Li-Network tsa Pele ho bona litlhaselo tsa bahanyetsi le ho fana ka tlhahiso ea mofuta o akaretsang oa koetliso ea bahanyetsi. Ho bontšoa hore kaho ea litlhaselo tse atlehileng tsa li-whitebox, tse amang ho bolela esale pele le ho qoba ho fumanoa, khahlano le Li-Network tsa Pele tse koetliselitsoeng ho CIFAR-10 le CIFAR-100 ho sebelisa mokhoa o reriloeng ho hloka boiteko bo boholo ho feta khahlanong le marang-rang a sirelelitsoeng a sebelisa mohanyetsi ea tloaelehileng. koetliso kapa MC-dropout.
- 21:10 Puisano ea sehlopha: "NeurlPS, e seng e holile haholo: ke mang ea molato le ho etsa eng?" - Alexander Krainov, Yandex
- 21:40 Mokete
Pherekhong 16, 18:30-21:30, Labone
- 19:00-19:30 "Ho rarolla mathata a ts'ebetso ho sebelisa R bakeng sa li-dummies" - Konstantin Firsov (Netris JSC, Moenjiniere ea ka Sehloohong oa Ts'ebetso).
- 19: 30-20: 00 "Optimization of inventory inventory" - Genrikh Ananyev (PJSC Beluga Group, Hlooho ea tlaleho ea automation).
- 20: 00-20: 30 "BMS ka X5: mokhoa oa ho etsa merafo ea ts'ebetso ea khoebo ho li-logs tsa POS tse sa etsoang ho sebelisa R" - Evgeniy Roldugin (X5 Retail Group, Hlooho ea Lefapha la Ts'ebeletso ea Ts'ebetso ea Ts'ebetso ea Lisebelisoa), Ilya Shutov (Media Tel, Hlooho oa ramahlale oa litaba oa Lefapha).
January 18, 12:00-18:00, Moqebelo
- "Ke neng moo ho leng bohlokoa ho ngola kopo ho tloha qalong, le mokhoa oa ho kholisa khoebo ka sena" - Alexey Pyzhyanov, moqapi, SiburPale ea 'nete ea kamoo re ileng ra sebetsana le likoloto tsa tekheniki ka tsela e feteletseng. Ke tla u bolella ka eona:
- Hobaneng ha kopo e ntle e fetohile lefa le lebe.
- Kamoo re entseng qeto e boima ea ho ngola tsohle bocha.
- Kamoo re rekisitseng mohopolo ona ho mong'a sehlahisoa.
- Ke eng e tsoileng khopolong ena qetellong, le hore na ke hobane’ng ha re sa ikoahlaele qeto eo re e entseng.
- "Vuejs API e soma" - Vladislav Prusov, moqapi oa Frontend, AGIMA
January 18, 12:00-15:00, Moqebelo
- 12:00 "Zindi Sendy Logistics Challenge (rus)" - Roman Pyankov
- 12:30 "Data Souls Wildfire AI (rus)" - Ilya Plotnikov
- 13:00 khefu ea Kofi
- 13:20 “Topcoder SpaceNet 5 Challenge & Signate The 3rd Tellus Satellite Challenge (eng)” - Ilya Kibardin
- 14:00 khefu ea Kofi
- 14:10 "Codalab Automated Time Series Regression (eng)" - Denis Vorotyntsev
Source: www.habr.com