Машинное обучение

Куратор секции: Татьяна Тюрина.

Здравствуйте. Меня зовут Татьяна Тюрина, я являюсь директором по развитию ИТ-компании RITG и куратором секции "Машинное обучение".

11 лет наша компания занимается разработкой высокотехнологичного программного обеспечения с использованием методов машинного обучения и нейронных сетей. С 2015 года мы принимаем участие в профильных конференциях и делимся опытом наших разработок среди аудитории единомышленников и представителей бизнеса. В этом году мы совместно с российскими экспертами в области работ с ИНС, Machine Learning, Data Mining расскажем аудитории секции о трендах данной области, про алгоритмы и трудности, которые Вас могут ожидать, а также о путях их решений. Слушатели узнают, что нужно начинающему обучателю нейронных сетей, чтобы сделать свой SkyNet; что такое нейроморфные архитектуры в вычислительных информационных моделях и как их применять. Впервые в регионе мы расскажем про "мозг на чипе" (обучение живой нейронной сети на мультиэлектродной матрице) и про создание вычислительной системы, близкой по своим возможностям к человеческому мозгу.

В общем, будет очень интересно и наглядно тем, кто уже применяет данные технологии и тем, кто только начал "погружаться". Ниже вы уже можете ознакомиться с нашими спикерами и кратким содержимым их выступлений.

Ждем Вас на "Стачке"!

Задать вопросы по секции:

E-mail: t.tyurina@ritg.ru

Facebook: t.tyurina

Alexey Natekin
Founder, Firekeeper @ DM Labs, Open Data Science

Many have heard about the impressive successes of Machine Learning and Data Science: while some technologies alter entire industries, others long ago unnoticeably became part of our lives.

However, here's the mishap: excessive attention of press and marketers to the success of common scientists and engineers has created so much information noise, that it became difficult understand what is what.

In this lecture, we will figure out what Machine Learning and Data Science are in general; learn how to ‘cook and eat’ them; and most importantly - what to do to join this feast of life and technology.

Nikita Zhiltsov
Analyst @ Rambler & Co

The report is devoted to modern text processing facilities based on machine learning, applied for some search tasks in Rambler & Co projects (portal, LJ). The speaker shares his experience in developing solutions based on the vector representation of word2vec and neural networks learning on real data. Examples of using the libraries of fastText, Keras and Tensorflow will be considered.

Alexander Krainov
Data Science Developer @ Restrim

1. What is special about direction of recommendations in the field of machine learning.

2. What are the demands to modern advisory systems and what has changed since the Netflix Prize competition.

3. Why all users are unhappy with the recommendations and how to improve it.

4. Difficulties and challenges in the development of an advisory system for IPTV nationwide.

5. What is the advantage of issuing recommendations in real time (online) mode and how to provide them with a minimum delay.

Michael Kiselev
CEO @ Megaputer Intelligence
  • Development of computational system with capacities approximated to those of the human brain.
  • Pulsed neural networks as physiologically similar models. Their peculiarities and advantages in comparison with traditional neural networks.
  • Prospective of pulsed neural networks implementation in form of specialized neuromorphic computers.
  • Key moment is learning algorithms for pulsed neural networks.
  • The most prospective approach is based on Hebb’s Law of synaptic plasticity. Algorithm L2TP as an example.
  • Synthesis of a network with necessary properties. Processes of network self-organization.

  • Sergey Voronov
    Senior development engineer @ RuGadget
    Senior development engineer
    • Development of robotic hand control system based on co-processing of EMG signal in the muscles responsible for functionality of hands and video information from cameras built in user's glasses.
    • Practical application of algorithms for movement adjustment, grasp type classification and motion start detection.

    Viktor Chernogorov
    Partner, Director for Development @ MobileUp
    St. Petersburg
    • Technology in start-ups. Examples of successful and failed cases in the last 2 years. Own and others'.
    • Is it possible to capitalize on innovations? How?
    • What is the outlook for new technologies?
    • What technologies will be implemented next?

    Sergey Lee
    "CyberCoach" (myosuit) project developer @ Biomechatronic Technologies Development Laboratory
    Nizhny Novgorod

    What is this?

    It is a general neuroelectronic interface – a system that reads the brain and muscles activity, and the body actions to be further sent to any actuators.

    The suit can be tailored to the unique properties of the human muscle groups, while sending recommendations regarding the workout mode to the AR goggles.

    You will witness: a practical demonstration of a myosuit by our engineer. The tablet will display data about recording the signal on the muscles. The basic operation principles will be demonstrated.

    Danil Pismenny
    Founder @ Brandy Mint

      We will give literacy classes on neural nets

      Answering the questions:

      • What tasks are already being solved with the help of machine learning and neural nets;
      • What instruments already exist and which of them are open;
      • What a beginning teacher of neural nets needs to make his/her own SkyNet.

    Dmitry Ermishin
    Chief IT analyst @ Finteсhlab

    Are there any new ways of multifactor authorization?

    Life goes digital. How to make it safer?

    Voice as a physical sign of personality identification.

    Why machine learning is that necessary.

    Sergey Stasenko
    Acting Head of Machine Learning and Data Analysis Laboratory @ Nizhny Novgorod State University
    Nizhny Novgorod
    • SPF vivarium: genetic engineering procedures for research purposes.
    • Control and coordination of muscle contractions.
    • Applying neuromorphic architecture to computational information models (in silico system).
    • Brain on a chip.
    • Living neural network learning on a multi-electrode array.

    Andrey Igonin
    CEO @ RITG

      Since we are IT professionals, rather than ultrasound specialists, we went the other way. Want to learn which? Then we are looking forward to seeing you at Stachka!

      We will tell you about the case that led to 150 time increase in customer’s productivity .

      Will tell about approaches to the use of classifiers ARTMAP and GPU with parallelization of computations.

      We will talk about the barriers that we faced and share experiences how to solve this problems.

    Pavel Erofeyev
    Lead Data Analysis Executive @ Airbus Group Innovations sk

    The airplane is a complex technical product characterized by continuous operation and life cycle exceeding 30 years.

    Modern airplanes come equipped with sophisticated mutually integrated systems comprising thousands of sensors of different types which generate increasingly fast data streams.

    With new sources of data and modern advancements in artificial intelligence, brand new approaches emerge to seemingly well-known aircraft life cycle support objectives, such as optimized fleet utilization, scheduled maintenance inspections, servicing, routine and extraordinary maintenance and repairs, etc.

    With new data handling technology we are now able to dramatically change the passenger transportation industry and to expand the urban transport space from two-dimensional to three-dimensional.

    However, development in this area is associated with conceptually new challenges concerning navigation and interactions among a great variety of agents.

    From the presentation, you will learn about the future of mass passenger transportation as seen by Airbus and get an insight to new data analysis and artificial intelligence challenges that are being tackled.