Technosphere Mail.Ru - three years

Today it is three years since the launch of one of our educational projects, the Mail.Ru Technosphere , implemented jointly with the Faculty of the VMK of Moscow State University. Lomonosov. The Technosphere program is designed to train specialists in the field of big data. Initially, it was designed for one year and consisted of six disciplines. However, a year later, we revised the program and made it biennial. During the four semesters, students study 12 disciplines, performing a large amount of practical work. At the same time, a preparatory course “Algorithms and Data Structures” was developed.

In the Technosphere take students 2-4 courses. Despite the fact that the scheme of entrance examinations in all our educational projects is the same (students take an online test and pass a face-to-face interview), in the Technosphere we focus more on the basic knowledge of higher mathematics. In addition to giving lectures, we created a laboratory where students work with real-life tasks that we encounter at Mail.Ru Group. For example, they are trying to improve analytical algorithms, to create certain heuristics. That is, they are doing the same thing that they would have done during a regular internship at the company. Since the fall of 2015, research began in the laboratory. For example, the possibilities of using neural networks for solving various business problems are being studied.

And in honor of the birthday, we post a list of training materials that are recommended for the study of our students throughout the two-year course.

Course: Algorithms for intelligent processing of large amounts of data

')
Lectures, articles and other materials

Mathematical monk blog
Large-Scale High-Precision Topic Modeling on Twitter
Comparison of cluster algorithms
Carnegie Mellon Lectures on Statistics: Data Mining, Lecture 22
Selection of materials on IPython and Jupyter
Visual Information Theory article on the theory of information
Sergey Nikolenko's blog on Habré
Learning representations by back propagation errors
Ruslan Salakhutdinov - Deep Learning
Yahoo! Hadoop tutorial
Tentative NumPy Tutorial
100 NumPy exercises
CRISP-DM User Guide
IBM CRISP-DM Guide
MapReduce: Simpli ﬁ ed Data Processing on Large Clusters

Literature

Pattern Recognition and Machine Learning
Here algorithms of inference are described, allowing to get quick approximate answers when it is permissible in specific situations. Graphic models are used to describe the probability distribution.

Data Mining Practical Machine Learning Tools and Techniques
The book describes the concepts of machine learning and gives practical advice on the use of tools and techniques in real data analysis tasks.

Introduction to Information Retrieval
The book teaches how to efficiently extract information using online search, text classification and clustering. It covers all aspects of the design and implementation of systems for collecting, indexing and searching for documents, methods for creating developing systems, the use of machine learning for working with text collections.

Mining of massive datasets
Here, the emphasis is on practical algorithms used to solve key problems in data analysis. The authors explain the various tricks associated with hashing, sensitive to locality, and with algorithms for processing fast incoming data. Also touched on the issues of the web, search for frequent sets of objects and clustering.

Pattern Classification
The book is devoted to neural networks, statistical pattern recognition, machine learning theory and invariance theory. Practical examples and comparisons of different methods are also given.

Machine Learning: a Probabilistic Perspective
The book is an introduction to machine learning based on a unified probabilistic approach.

An Introduction to Data Science
A book for those who make the first steps in data processing. Code samples on R are presented for solving various interesting problems.

Data Mining and Knowledge Discovery Handbook
It describes key ideas, theories, standards, methodologies, trends, complexities, and methods for applying in-depth data analysis.

Stochastic Gradient Descent Tricks
This is the first chapter of the book “Neural Networks, Tricks of the Trade”, it discusses the method of stochastic back-propagation for training neural networks. In essence, this is a type of stochastic gradient descent technique.

Neural Networks and Learning Machines (3rd Edition)
The book discusses the modern methods of application of neural networks from an engineering point of view. Matlab code samples can be downloaded from here .

The elements of statistical learning
It describes important ideas in the areas of statistics, data processing, machine learning and bioinformatics.

Ensemble Methods: Foundations and Algorithms
The book describes the theory and algorithms of machine learning, from simple to more complex.