Writing a Ph.D.

    This post may be useful to graduate students (and masters) when writing their scientific papers, as it contains some observations and conclusions made by the author while working on the Ph.D.

    Topic selection


    Probably the most important point and also the most difficult. There can be many reasons - from organizational ones, such as a change of university, department, leader to quite ordinary ones - the current direction is not interesting, but a new one has not yet been invented. But, as a rule, there is only one problem - which topic to choose for the candidate?

    A few tips:
    • if you can’t decide, but the topic name is already required from you, then first select the neutral type “Modeling complex objects in the face of uncertainty”, which covers 80% of modern objects, etc. After a while, you can change the topic, add specifics, but formally the changes will not be dramatic and, in theory, there should be no problems;
    • convince your supervisor that you don’t need to choose a topic that he knows well, not you. Although he should understand the topic that you offer him;
    • and probably the most important thing: if you don’t know what to choose, ask yourself a simple question: “what do I know best?” or "and if they gave me a lot of money for research, what would I like to do?" I think you will find the answer quickly (if not, then maybe you should not go to graduate school?).

    Work plan


    I was always told and persuaded that the plan should always be at the very beginning of work in order to understand where we are now. Frankly, only the name “Plan” remains from the original plan. Many here may not agree, they say, if you can’t even make a plan, then what can you even do? I will not argue, perhaps someone else, but in practice I did not notice this.

    Therefore, it is better to draw up a general plan, which will then be amended according to your research and results.

    Start research


    Many people before entering graduate school (and even after that) fall into a state that is close to fear and anxiety. Everyone is tormented by questions: “so much has already been done in my field, what can I bring new and unique?”, “A and B did this back in the 60s and wrote 10 books each, how will I compete with them?

    First of all, you need to understand that not all in this world have been invented and far from everything invented is optimal. Therefore, sit down at a computer and act on the principle:
    • problem -> is it solved? -> yes -> new problem ->
    • problem -> is it solved? -> no or bad -> solution options -> research -> results
    • problem -> is it solved? -> yes -> optimal solutions -> no -> improved solution options -> research -> results
    Signs that your topic is relevant:
    • they talk about your topic as tomorrow
    • the literature on your subject is in its infancy
    • Russian / Ukrainian literature, no scientific articles
    • The scientific laboratories of leading companies are involved in your topic.
    • your topic relates to a rapidly changing field (information retrieval, object recognition)
    I want to immediately note that the absence of any mention (work, research. Articles) on your topic does not mean that your topic is relevant, but that nobody needs it, since nobody raised it anywhere.

    Scientific novelty


    What can be considered scientific novelty?

    So this is:
    • ideally: a new model, a new method
    • improvement of the model method
    • application (adaptation) of a known method, model to new areas
    • improving the quality, accuracy, speed, number of operations, etc. (it is necessary to specify and justify quality criteria)
    • new or improved methodology or approach.
    Important! The algorithm is not considered scientific novelty, but is only a practical implementation of your model or method.

    One whole


    Your work should be a single whole, i.e. all your results should be logically related. It seems obvious, but in practice it’s more complicated. As a rule, you get different results, use different methods from different areas, and then it’s very difficult (I don’t say impossible) to tie it all together. For example, I had to associate methods for obtaining associative rules, clustering methods, classical statistical methods, information theory and semantics. It is even more difficult when you studied one subject in the magistracy, and then switched to another. In this case, all of your old scientific publications may not fit into the new topic. And if you have a shortage of work and you want to tie them, then it will be difficult to do.

    Scientific publications, patents, acts of implementation


    Well, that one is really simple - they are needed and as much as possible :)

    An important fact is that you should have published articles on all points of scientific novelty. This is an important requirement and, as a rule, brings its troubles at the end of the work.

    That's all for now, I hope that this topic will help someone. If there are any clarifying questions or are interested in certain issues (for example, acts of implementation or scientific articles), write, I will try to help.

    Also popular now: