Faculty of Physics launches Neural Networks Course

Inspired by the idea of giving the more promising graduate students at Lomonosov Moscow State University an opportunity to gain world-class empirical expertise in neural networks for data mining use in their research, the developer team set to work at the end of 2020. The outcome of their work, the study course “Neural Networks and Their Applications in Research”, was supported and launched by the Intellect Foundation.

The course became part of the variable graduate school curriculum at the Faculty of Physics as of February 2021, offered as a specialist study course.

The great thing about this course is that it welcomes young scholars from among science and liberal arts students, not just those with engineering majors. In fact, the course creators are positive that machine learning, and specifically neural networks, are eminently applicable to research in vastly different fields. The key prerequisites for course entrants are programming skills and a thesis subject germane to the use of neural network methods.

The course is 2 semesters long.

The first semester includes 15 classes, taught once a week in each group. The instructors on the course present practical material on neural network programming in the Python language using the PyTorch library. Every class consists of a theoretical part (lecture) and a practical part (programming assignment), and is planned in JupyterNotebook format. The lecture presents several sample codes to illustrate the topic just learned. The instructor launches the samples in a GoogleColab environment to provide a graphic explanation on how one or another neural network model works.

When they are through with the theory part, the graduate students are asked to complete a series of independent assignments, which are then graded by the instructor.

The first six lectures of the course are up on the course's YouTube channel..

In the course's second semester, the students are supposed to write their course papers and research articles, supervised by the instructor acting as a second research advisor to the student. The idea is that the group's instructor will study the student's specific dataset and research problem and assist them in the mining for the right approach to solving it (the student is to mine the approaches independently) by offering specialist methods, and will eventually review and critique the outputs. The use of neural networks in research should not be the goal in and of itself. The work is to yield such scientific, critically assayed outputs as would not be possible without the use of machine learning methods.

For the purposes of distance learning, the course “Neural Networks and Their Applications in Research” is also projected as an online course available to all those wishing to learn.

And last but not least, the Intellect Foundation offers financial support for students on the course, which includes a scholarship program and rewards for published coursework.