Nā hanana me ka hoʻopaʻa inoa hāmama:
Ianuali 14, 19:00-22:00, Poalua
Ke kono nei mākou iā ʻoe i kahi hālāwai e pili ana i ka naʻauao artificial, kāna noi ma nā polokalamu kelepona a me nā ʻano ʻenehana a me nā ʻoihana koʻikoʻi o ka makahiki hou. Aia ka papahana i nā hōʻike hoihoi, nā kūkākūkā, ka pizza a me ke ʻano maikaʻi.
ʻO kekahi o nā mea haʻi'ōlelo he paionia i ka hoʻokomoʻana i nāʻenehana hou loa ma Hollywood, ka Hale Paʻi; ʻO kāna puke "Augmented: Life in the Smart Lane" i ʻōlelo ʻia ʻo ia kekahi o kāna puke punahele punahele e ka Pelekikena o Kina ma kāna haʻiʻōlelo makahiki hou.
Ianuali 15, e hoomaka ana ma ka hora 18:00, Poakolu
- 18:00 Kakau inoa
- 19:00 Wehe - Mikhail Bilenko, Yandex
- 19:05 Ke aʻo hou ʻana ma NeurIPS 2019: pehea ia - Sergey Kolesnikov, TinkoffI kēlā me kēia makahiki ke ulu nei ke kumuhana o ka hoʻoikaika ʻana i ke aʻo ʻana (RL). A i kēlā me kēia makahiki, hoʻohui ʻo DeepMind a me OpenAI i ka wahie i ke ahi ma ka hoʻokuʻu ʻana i kahi bot hana superhuman hou. Aia kekahi mea kūpono ma hope o kēia? A he aha nā ʻano hou loa i nā ʻokoʻa RL āpau? E ʻike kākou!
- 19:25 Nānā i ka hana NLP ma NeurIPS 2019 - Mikhail Burtsev, MIPTI kēia mau lā, ʻo nā ʻano holomua nui loa i ke kahua o ka hoʻoponopono ʻōlelo kūlohelohe e pili ana i ke kūkulu ʻana i nā hale kiʻi ma muli o nā hiʻohiʻona ʻōlelo a me nā kiʻi ʻike. Hāʻawi ka hōʻike i kahi ʻike o nā hana i hoʻohana ʻia ai kēia mau ʻano hana e kūkulu i nā ʻōnaehana kamaʻilio e hoʻokō i nā hana like ʻole. No ka laʻana, no ke kamaʻilio ʻana i nā kumuhana maʻamau, e hoʻonui i ka empathy a me ke alakaʻi ʻana i ke kamaʻilio e pili ana i ka pahuhopu.
- 19:45 Ala e hoʻomaopopo i ke ʻano o ka ʻili o ka hana poho - Dmitry Vetrov, Faculty of Computer Science, National Research University Higher School of EconomicsE kūkākūkā au i kekahi mau pepa e ʻimi ana i nā hopena maʻamau i ke aʻo hohonu. ʻO kēia mau hopena e hoʻomālamalama i ke ʻano o ka ʻili o ka hana pohō ma ke keʻena paona a hiki iā mākou ke waiho i mua i kekahi mau hypotheses. Inā hoʻopaʻa ʻia, hiki ke hoʻoponopono maikaʻi i ka nui o ka ʻanuʻu i nā ʻano loiloi. E hiki no hoi keia ke wanana i ka waiwai o ka hana poho ma ka laana hoao ma mua o ka pau ana o ke a'o.
- 20:05 Nānā i nā hana ma ka ʻike kamepiula ma NeurIPS 2019 - Sergey Ovcharenko, Konstantin Lakhman, YandexE nānā mākou i nā wahi nui o ka noiʻi a me ka hana ma ka ʻike kamepiula. E ho'āʻo kākou e hoʻomaopopo inā ua hoʻoholo ʻia nā pilikia a pau mai ka manaʻo o ke kula, inā paha e hoʻomau ana ka huakaʻi lanakila o GAN ma nā wahi āpau, ka mea e kūʻē nei, a i ka wā e hiki mai ai ke kipi i mālama ʻole ʻia.
- 20:25 LK ʻĒ
- 20:40 Ka hoʻohālike ʻana i nā kaʻina me ka palena ʻole o ka hanauna - Dmitry Emelianenko, YandexHāʻawi mākou i kahi kumu hoʻohālike e hiki ke hoʻokomo i nā huaʻōlelo i nā wahi kūʻokoʻa i ka huaʻōlelo i hana ʻia. Aʻo maoli ke kumu hoʻohālike i kahi kauoha decoding kūpono ma muli o ka ʻikepili. Loaʻa ka maikaʻi maikaʻi loa ma kekahi mau waihona: no ka unuhi mīkini, hoʻohana ma LaTeX a me ka wehewehe kiʻi. Hoʻolaʻa ʻia ka hōʻike i kahi ʻatikala a mākou e hōʻike ai i ke ʻano o ka hoʻoponopono decoding i aʻo ʻia i kūpono a kikoʻī i ka pilikia e hoʻoholo ʻia.
- 20:55 Reverse KL-Divergence Training of Prior Networks: Improved Uncertainty and Adversarial Robustness - Andrey Malinin, YandexUa hoʻohana ʻia nā ala hui no ka helu ʻana i ka maopopo ʻole i nā hana o ka ʻike misclassification, out-of-distribution input detection and adversarial attack detection. Ua noi ʻia ʻo Prior Networks ma ke ʻano he ala e hoʻohālikelike pono ai i kahi hui o nā hiʻohiʻona no ka hoʻokaʻawale ʻana ma o ka hoʻohālikelike ʻana i kahi Dirichlet ma mua o ka hāʻawi ʻana ma luna o ka puʻunaue puka. Ua hōʻike ʻia kēia mau hiʻohiʻona e ʻoi aku ka maikaʻi o nā ʻano hui ʻē aʻe, e like me Monte-Carlo Dropout, ma ka hana o ka ʻike komo ʻana ma waho. Eia nō naʻe, paʻakikī ka hoʻonui ʻana i nā Pūnaewele Mua i nā ʻikepili paʻakikī me nā papa he nui me ka hoʻohana ʻana i nā pae hoʻomaʻamaʻa i manaʻo mua ʻia. Hāʻawi kēia pepa i ʻelua mau haʻawina. ʻO ka mea mua, hōʻike mākou i ke kumu hoʻomaʻamaʻa kūpono no Prior Networks ka hoʻohuli KL-divergence ma waena o nā māhele Dirichlet. Hoʻopuka kēia mau pilikia i ke ʻano o ka puʻunaue ʻikepili hoʻomaʻamaʻa, hiki i nā pūnaewele mua ke hoʻomaʻamaʻa maikaʻi ʻia i nā hana hoʻokaʻawale me nā papa he nui, a me ka hoʻomaikaʻi ʻana i ka hana ʻike ma waho. ʻO ka lua, me ka hoʻohana ʻana i kēia ʻano hoʻomaʻamaʻa hou, e noiʻi ana kēia pepa me ka hoʻohana ʻana i Prior Networks no ka ʻike ʻana i nā hoʻouka kaua ʻana a hoʻopuka i kahi ʻano ākea o ka hoʻomaʻamaʻa ʻenemi. Ua hōʻike ʻia ʻo ke kūkulu ʻana i ka hoʻouka ʻana o ka pahu keʻokeʻo adaptive kūleʻa, e pili ana i ka wānana a me ka pale ʻana i ka ʻike ʻana, e kūʻē i ka Prior Networks i hoʻomaʻamaʻa ʻia ma CIFAR-10 a me CIFAR-100 me ka hoʻohana ʻana i ke ala i manaʻo ʻia e koi i ka nui o ka hoʻoikaika ʻana ma mua o nā ʻoihana i pale ʻia me ka hoʻohana ʻana i ka ʻenemi maʻamau. hoʻomaʻamaʻa a i ʻole MC-haʻalele.
- 21:10 Kūkākūkā Panel: "NeurlPS, ka mea i ulu nui loa: ʻo wai ka hewa a he aha ka hana?" - Alexander Krainov, Yandex
- 21:40 LK Ma hope o ka pāʻina
Ianuali 16, 18:30-21:30, Poaha
- 19:00-19:30 "Hoʻoholo i nā pilikia hana me ka hoʻohana ʻana i R no nā dummies" - Konstantin Firsov (Netris JSC, Luna Hoʻokō Nui).
- 19:30-20:00 "Optimization o ka waihona waiwai ma ke kūʻai aku" - Genrikh Ananyev (PJSC Beluga Group, Poʻo o ka hōʻike automation).
- 20: 00-20: 30 "BMS ma X5: pehea e hana ai i ka hana ʻoihana ʻoihana ma nā lāʻau POS i kūkulu ʻole ʻia me ka hoʻohana ʻana iā R" - Evgeniy Roldugin (X5 Retail Group, Head of Service Quality Control Tools Department), Ilya Shutov (Media Tel, Head o ka 'epekema data 'Oihana).
Ianuali 18, 12:00-18:00, Poaono
- "I ka manawa hea e pono ai ke kākau hou ʻana i kahi noi mai ka wā ʻōpala, a pehea e hōʻoiaʻiʻo ai i ka ʻoihana o kēia" - Alexey Pyzhyanov, mea hoʻomohala, SiburʻO ka moʻolelo maoli o ke ʻano o kā mākou hana ʻana i ka aie ʻenehana ma ke ʻano radical loa. E haʻi aku wau iā ʻoe no ia mea:
- No ke aha i lilo ai kahi noi maikaʻi i hoʻoilina weliweli.
- Pehea mākou i hoʻoholo ai e kākau hou i nā mea a pau.
- Pehea mākou i kūʻai aku ai i kēia manaʻo i ka mea nona ka huahana.
- He aha ka mea i puka mai i kēia manaʻo i ka hopena, a no ke aha mākou e mihi ʻole ai i ka hoʻoholo a mākou i hana ai.
- "Hoʻohenehene ʻo Vuejs API" - Vladislav Prusov, mea hoʻomohala Frontend, AGIMA
Ianuali 18, 12:00-15:00, Poaono
- 12:00 "Zindi Sendy Logistics Challenge (rus)" - Roman Pyankov
- 12:30 "Data Souls Wildfire AI (rus)" - Ilya Plotnikov
- 13:00 Hoʻomaha kofe
- 13:20 "Topcoder SpaceNet 5 Challenge & Signing The 3rd Tellus Satellite Challenge (eng)" - Ilya Kibardin
- 14:00 Hoʻomaha kofe
- 14:10 "Codalab Automated Time Series Regression (eng)" - Denis Vorotyntsev
Source: www.habr.com