The recommendations below on the conduct of research work are addressed to students and can be useful at all stages of training, from the choice of the supervisor to the preparation of the final qualification work.

First steps

Working with a supervisor usually begins with the solution of several test problems and / or reading several articles on the topic of future scientific work, possibly in English. Then you will be given the main task, which in the future should grow into the theme of the final work. After receiving the next assignment, do not hesitate to ask for additional explanations. Much worse, if you are dug for a long time to disappear, so nothing is done.
Another common mistake is to postpone scientific work for later. Usually the manager expects that your work will join the general research and expects certain results by certain dates. If you cope with the first task quickly, then you get complication, then the next, and by the end of the study, tangible results run in. It is practically impossible to do something worthwhile for a couple of weeks. Not surprisingly, if such a course or graduation work will be evaluated not higher than the top three. Good ideas appear as a result of repeated joint discussions, and not immediately. It takes time to understand the reasons for the failures of the first experiments, to come up with a better algorithm, to understand something and to prove its properties. Therefore, a good course work can be done objectively only in a couple of semesters. Your manager has the right to be busy, do not take the time to read the material you sent for a whole week (two, three, …), once again postpone the meeting for a week later, and provide other signs of inattention. This is normal. Spend the time to understand on your own what to do next or do self-education. No one will give the student a topic on which there is not a single publication in the world. Do not forget about the search in Google, Google Scholar, arXiv, CiteSeer, Sci-Hub. The name of the theory, or algorithm, or formulas that you discussed with the supervisor are already keywords for searching. Find the sites of top scientific conferences in your area and find out what problems the scientific community is currently engaged in. Be prepared for the fact that even a superficial acquaintance will take a lot of time, and much will not be clear. But you will have the right questions.

Do not be afraid to make mistakes.
You are learning a complicated business, so no one will judge you if it does not happen all at once. Remember that the scientific community values creativity, hard work and a desire for knowledge. It is generally very democratic and very friendly. If your supervisor was not that, change it sooner. Psychological discomfort from communication for some people is a strong demotivator and a brake in the work. The scientific problem is never completely understood. The supervisor can understand a little more than you. However, he may not know in detail how to solve it. You need it for this. There is a simple algorithm, how not to be stupid and not to stagnate on the spot in the conditions of uncertainty of the job. Write a complete list of questions that you do not understand in order to reach the goal. Write three options for your answers to each question. Discuss them with the supervisor. The conversation “which of the three options is better” is always more constructive than the conversation “think up for yourself how to do it all”. Choose for each question of the three options the easiest in terms of implementation. Complete the entire task from beginning to end, so that at least something will work. If this is a data analysis task, then it is necessary to reach the evaluation of the quality of solutions. Discuss the first results with the supervisor. Gradually, one by one, replace the patches with more reasonable options and see what improves. Gradually, one by one, replace the patches with more reasonable options and see what improves. This is a universal block diagram for those who like full clarity before work, and otherwise will doubt indefinitely (print large and hang on the wall). Perhaps you are not that, and you will personally need another algorithm of success. There are no universal recipes for everyone.