7

I employ an incremental approach for my PhD research. In other words, at first, I propose a method to solve a problem and publish a paper about it. Then, I update my method by either improving its performance/optimality or adding more features so that it can handle more cases and I keep publishing papers reporting my progress. As a result, I have a pretty good publishing record, more than 5 papers as the first author in less than 3 years. However, between each paper, I sometimes make some major improvement such as completely replacing an old approach by a new (and better) one to solve the same problem.

At the moment, I am about to start writing up and face the problem of how to include all my papers in the thesis. My initial plan is to mainly focus on the last paper that I have published since it contains the best version of my research. However, doing so potentially makes my PhD research look weak since I have published many papers but only use one as the major theme of my thesis.

On the other hand, I am thinking about writing my thesis as a research diary that reports my journey to investigate how to solve one problem, e.g. at first I start with this approach, then after further research, I employ another one which results in better performance... But I am not sure if it is a right way to write my thesis.

My question is how to write a thesis for a incremental research. It would be great to hear from someone who was in the same situation. My field is computer science but answers from other fields are welcome.

  • 4
    Why don't you ask your thesis adviser? It's their job to answer question like this and they are familiar with the customs in your country, scientific field, department .... In Germany those publications would be part of your thesis and you would write only about 15 to 20 pages putting them into context. – Roland May 9 '16 at 10:32
  • 2
    @Roland I am already discussing with my supervisor, I just want to have extra information before making the final decision :) – Long Thai May 9 '16 at 11:03
  • 5
    A thesis should be a story...and "The development of a method for X" seems like a good story. Rather than focusing only on the final version, your methods can discuss how your incremental approach developed and the insights that led to the improvements – user24098 May 9 '16 at 14:18
  • @Roland Not all examination regulations in Germany allow to staple together your papers and write a few pages for putting them into context. Some still insist on a classical thesis format. – DCTLib May 10 '16 at 7:59
3

From the point of view of the reader, jumping to the most sophisticated version of a method can be difficult. A thesis isn't a textbook, but there are parallels. A textbook often uses an incremental approach starting with almost a toy model to get the basic concepts sorted, before making it more realistic. The assumed background knowledge of someone reading your thesis might let you skip the first step but not necessarily (don't write just for the examiners). It also allows you to indicate why you made certain decisions in the final version.

2

So essentially, you went about your doctorate in the manner a tenured prof goes about a problem: milk it thoroughly in a sequence of incremental results. If your papers are indeed incremental, without trade-offs that call of deep analysis, I don't think this makes for a great doctoral work. And I'm speaking from a CS perspective. And I would like to caution any current/prospective students reading this. Unless this is a monumental problem, what do you think is gained out of presenting 5 different ways of solving it (as a doctoral thesis)? You ought to have moved to neighbouring problems, or to a completely different one, rather than climbing up the same hill over and over again. But what is past is past.

I would take the presentation of such a thesis very seriously. Maybe you will need to work a little harder to put all of these solutions in perspective. Maybe this requires more work. But think properly about "what is to be gained out of presenting all of these different ways?" Clearly, you have done the problem to death by now. It seems to me that your best chance is to analyse the various solutions in an intelligent manner.

1

Your research is very interesting (I like your approach lol). It seems like you started with specific algorithm or solution to solve the problem and then you make it more and more general and/or optimize it so it can cover more cases. From one solution you discuss its limitation and problem and then propose a new, modified approach/optimization from the first one. So that will be a fluid story line from one paper to another. If I'm your thesis reader, I would like to see that kind of story telling :)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.