About

This is the extension of “Automation of Scientific Research” course. The work is organized in teams (2-3 people). Students will prepare weekly presentation of the team’s work: 5 minutes presentation + 5 minutes discussion with other students.

Criteria for the project:

1. The entire project must be on GitHub under an OpenSource license (MIT)
2. The project must contain the code and instructions for launching it.
   1. The basic code in jupyter notebook to demonstrate the concept of work and plotting from the article should be run with colab.
   2. The source code of the computational experiment should be run on a Unix system by executing two commands (possibly in a special docker image):
      • python3 train.py - for training models.
      • python3 test.py - for testing, obtaining the results of the experiment.
   3. The documentation for the code is sphinx.
3. The project should contain the manuscript of the article using a stylistic from arXiv - for unification.
4. If the project implies non-synthetic data, then there should be an instruction for obtaining this data, as well as a script for obtaining them. If the data is specific, then they need to be posted on one of the file storages.

Syllabus

• Introductory lecture.
• Projects discussion.
• Project structure.
• The first checkpoint.
• Paper structure.
• How to build a jupiter notebook correctly.
• The second checkpoint.
• The correct code is not in jupiter notebook!
• Data storage for an experiment.
• The third checkpoint.
• Code documentation using sphinx.
• Docker for the experiment code.
• The fourth checkpoint.
• Evaluation.

Labworks

Writing scientific paper.

Grading

There are 4 checkpoints in the course, 3 points are given for each

1. Research of the subject area - analysis of the problem.
2. Theoretical results.
3. Computational experiment.
4. Github project.

Prerequisites

Machine learning, deep learning, Python.