Faculty Graduates Participated in the Microsoft Research Summer School ‘Machine Learning and Intellect’
Boris Kovalenko, Stanislav Semenov, and Alexandra Fenster, 2015 graduates of the master’s programme ‘Data Science’, took part in the event.
The summer school was targeted at senior students, doctoral students, and young researchers. During the week the participants learned about the key tasks of the machine-learning era and took part in a hackathon.
Anton Konushin, Academic Director of the educational programme ‘Applied Mathematics and Information Science’, who has been part of the School’s organizing committee several times, also participated in it this time as well. He spoke on Machine Learning for Videos, while other lecturers at the event included Dmitry Vetrov, Head of the Big Data and Information Retrieval School, who spoke on Bayesian Inference and Latent Variable Models in Machine Learning, and Michael Levin, Lecturer at the Joint Department with Yandex, who spoke on MatrixNet and Applications.
Academic Director of the educational programme ‘Applied Mathematics and Information Science’
Machine learning is one of the fastest developing areas of computer science, so the huge interest in this school was no surprise. And many people feel that over the last couple of years, progress in this area has gained momentum. For example, Ruslan Salakhutdinov and Kate Saenko told us about successes in automatically constructing text descriptions for videos, which has been achieved by means of recurrent neural networks linking visual and textual information as well as taking into account language grammar. In some cases, it is becoming difficult to distinguish a text written by a human from a text generated by a computer.
I learned about the school from the Habrahabrwebsite. It was very interesting to listen to the introductory lectures by Christopher Bishop (Microsoft Research, UK) on Model-Based Machine Learning, and Ruslan Salakhutdinov ( University of Toronto, Canada) on Deep Learning, as well as by Dmitry Vetrov and Anton Konushin. Industry professionals from Yandex, nVidia, and Microsoft gave interesting talks on various tools and products.
In order to participate in the project, it was necessary to submit a portfolio with term papers, graduate papers and articles, as well as university grades etc. According to the organizers, 70 participants were selected out of 700 applicants, meaning that there were 10 applicants for each place.
This school was pretty intensive. In addition to interesting lectures from leading professionals, there was also hands-on experience. An AutoML Hackathon took place as part of the school, and the competition included the creation of a ‘Perfect Blackbox’, which would make predictions for a binary classification problem in the AutoML database on datasets not described before and ideally on any datasets. I was particularly pleased that participants were split into teams by the hackathon organizers. Thanks to this, I got experience of working with wonderful and very strong teammates, many of whom I didn’t know before the hackathon. Communication was a very important aspect of the school. It was also particularly important for me that the topic of Deep Learning in images and video was presented in detail. As a result, I was able to get answers to some urgent practical questions.
Общение вообще было очень важной стороной школы. Для меня особенно важно было то, что Deep Learning в изображениях и видео был представлен как топик в большом объёме. В результате я смогла получить ответы на свои насущные практические вопросы.
A hackathon on machine learning took place during the school. We needed to code a program that would receive a teaching sample – information on objects (numerical and categorical indicators), and automatically build a system for predicting a target indicator within a limited period of time. My team managed to take first place, since we got the best prediction result on the test sample. I’m quite closely involved in machine learning. For example, I’m one of the top five in the world in terms of solving applied tasks of data analysis and machine learning according to the kaggle ranking and I’m a professor of machine learning at the Yandex Data Analysis School.
On September 13, 2015, on Programmer’s Day, the Faculty of Computer Science is organizing an event for school students. IT experts and researchers will speak on today’s machine learning application in high energy physics, computer vision, the links between IT technologies and Lev Tolstoy’s writings, and other development on the borders of IT, mathematics, Russian language, and biology. The programme will also include master classes and contests. Register for the event.