Open course "Deep Learning on the fingers"

    After February 18th, the open and free “Deep Learning on Fingers” course will begin.

    The course is designed to deal with modern deep learning from scratch, and does not require knowledge of either neural networks or machine learning in general. Lectures with streams on Youtube, assignments on Python, discussions and assistance in the best Russian-speaking DS communities - and ClosedCircles .

    After it, you will not become an expert, but you will understand that all this, you will be able to apply DL in practice and will be able to understand further yourself. Well, at best.

    At the same time and in the same volume, the course will be read for undergraduates of the Novosibirsk State University, as well as students of the CS Center of Novosibirsk.

    The explanation on the fingers will look like this:

    The main link is . Details below.

    Where did it all come from

    It all started with the fact that one of the former scientific leaders once again came to visit, and I asked how they were doing with machine learning in the native department of the NSU. It turned out that they had been looking for a teacher for several years and could not find it. I jokingly said that in principle I can, but I will have to arrange a teleconference to Novosibirsk from Silicon Valley.

    Word for word, in the next academic year, students of the department came to the audience, and they were included on the Youtube floodlight stream. They asked questions, I answered with a 5-second delay, and in general, the future has come. Many thanks to the department of AFTI FF NSU and Yura Baburov ( buriy ), who led the seminars!

    Well, if you do in 2019 a university course on enthusiasm, then it is necessary that its materials are open and accessible to all.

    Moreover, when I started doing this, there were already excellent courses in English about Deep Learning - both CS231n from Stanford and Andrej Karpathy, and from Jeremy Howard. About classic ML in Russian there is an excellent from the yorko and ODS brethren, but I did not find about the DL in Russian a complete set (lectures, tasks and a lively place for discussion).

    Now, by the way, many are doing and posting Deep Learning courses, and that's great. If you manage to be at least some time a baseline for other courses - for me this is already a success.

    What's in the know

    What do you need for a full course? For me - lectures, assignments, and a place for discussions and questions.

    Lectures - will stream on Youtube every week so that you can ask questions on the go and then.
    Tasks - will be laid out on github, all on Python in laptops.
    To perform tasks to have your own GPU is optional - thanks to Google for Google Colab !

    The discussion site is the #dlcourse_ai channel in and #dlcourse on ClosedCircles.

    We are also in the course of asking you to write a post about some modern article and take part in the actual machine learning competition.

    How to get into ODS

    Доступ в ODS дается по инвайтам, но если вы в заявке укажете, что вы насчет курса, ваши шансы резко возрастут. Одновременно это уверенно отвечает на вопрос "чем этот курс лучше других".

    A sample program can be viewed in the version of last year, about which there was a post on Habré -

    TL; DR

    • Fundamentals of deep learning to formulas, mathematics and the main method of training.
    • Using modern libraries on the example of PyTorch.
    • Application to the main domains - computer vision, word processing and sound, reinforcement training.
    • Overview of the current state.

    What you need to know for a successful course

    Three things!
    For each, a verification picture will be attached - if it did not frighten you, then you are in.

    Some python:

    A bit of linear algebra:

    A bit of mathematical analysis:

    If any of the pictures are still scary, here is a good list of options for what to do with it - .

    Who leads the course?

    Are you really a producer? Yes, I really am a producer.

    Lectures will be delivered mainly by me, but guest lectures from well-known specialists in narrow circles are also planned.
    Just in case, two words about me - my name is Semyon Kozlov, I live in the Valley, I manage machine learning in the Instrumental startup , before I worked in the machine learning team at Dropbox. Earlier, I was engaged in computer graphics, made games, and helped deep learning by giving video cards a reason to exist.

    With the rest help:

    • Yuri Baburov, a specialist in text analysis and speech recognition
    • Pavel Petrochenko and Denis Denisenko from OnPositive
    • Maxim Vakhrushev and Kirill Broadt from Novosibirsk CS-Center
    • Alexander Goncharenko from Expasoft
    • Lena Bruches from 2gis
    • Guest lecturers - until I get to blame!
    • All ODS tusovka helping to collect lecture material.

    I also have a horror as my wife helps, supporting and freeing time. Cooking all this is some crazy amount of effort.

    How to start!

    At you can sign up for anos in the mail or in the Telegram-group.
    The first lecture is two weeks later, after February 18th.
    If you have long been going to deal with this all neural networks - here it is, a chance!

    And also - if everything works out for us, then inevitably all courses of all universities will be conducted exactly like this. Let's fight for a wonderful future!

    Also popular now: