yandex_2017 - Training
CSC's trainings and events have moved
Find our upcoming trainings and events at www.csc.fi.
This site is an archive version and is no longer updated.
Date: | 02.02.2017 9:00 - 04.02.2017 17:00 |
Location details: | The event is organised at the CSC Training Facilities located in the premises of CSC at Keilaranta 14, Espoo, Finland. The best way to reach us is by public transportation; more detailed travel tips are available. |
Language: | english-language |
lecturers: |
Andrey Ustyuzhanin (Yandex) Maxim Borisyak (Yandex) Mikhail Usvyatsov (Yandex) Alexander Panin (Yandex) |
Price: |
|
The fee covers all materials, lunches as well as morning and afternoon coffees. |
Practicalities: event-support@csc.fi
Description
This course is organized by Yandex School of Data Analysis, Higher School of Economics (hse.ru), and CSC and gives an introduction to deep learning, convolutional and recurrent neural networks, GPU computing, and commonly used tools to train and apply deep neural networks for various applications.
The course consists of lectures and hands-on exercises. The main tools in the exercise will be Theano and Lasagne. CSC's Taito-GPU environment will be used in the exercise sessions.
Learning outcome
After the course the participants should have the skills needed for applying deep learning for different tasks and utilizing the available GPU capacity at CSC for training and deploying their own neural networks.
Prerequisities
The participants are assumed to have working knowledge of Python and suitable background in data analysis or machine learning. Note that this is a more advanced course and fundamentals of machine learning are not covered on this course. In addition, fluent operation in a Linux/Unix environment will be assumed.
Materials:
Setup: https://github.com/CSC-IT-Center-for-Science/machine-learning-scripts/tree/master/courses/yandex2017
Course repository: https://github.com/yandexdataschool/CSC_deeplearning
Day 1: Introduction to deep learning and tools
9:00 - 10:30 Lecture: Introduction, linear models, stochastic gradient descent
10:30 - 12:00 Seminar
12:00 - 13:30 Lunch
13:30 - 15:00 Lecture: Basic deep / representation learning, backpropagation, initialization, philosophy
15:00 - 16:30 Seminar
16:30 - 18:00 Extra: Yandex intro
Day 2: Convolutional Neural Networks
9:00 - 10:30 Lecture: Computer vision, convolutional neural networks (CNN), batch normalization, data augmentation
10:30 - 12:00 Seminar
12:00 - 13:30 Lunch
13:30 - 15:00 Lecture: Natural language processing, embedding, text CNN
15:00 - 16:30 Seminar
16:30 - 18:00 Extra: Generative adversial networks (GAN)
Day 3: Recurrent Neural Networks
9:00 - 10:30 Lecture: Recurrent neural nets, backpropagation through time, gradient explosion/vanishing
10:30 - 12:00 Seminar
12:00 - 13:30 Lunch
13:30 - 15:00 Lecture: Captioning
15:00 - 16:30 Seminar
16:30 - 18:00 Extra: Cool use cases