Finally, the most important thing that Professor Buchanan suggested was that there all theses can be reduced to six chapters ... no more, no less. I've generally found this to be true, so I give you the six chapters.
|1) Introduction||what's the problem you're trying to solve?
|2) Extended Example||give an example of your system and how it works. i stress the word example. use the chapter to give a hypothetical (better yet, a real) scenario of use. Most ML theses are application oriented, so this is easy. For those of you who aren't doing applications, we should talk about how to make this fly.|
|3) Theory/Rationale||why are you doing what you're doing? who are the giants whose shoulders you stand on? this is the opportunity to get into detailed surveys of previous work. but, this is not meant to be a literature review. this chapter should connect prior work to your own; it should justify the implementation you'll describe in the next chapter.|
|4) Design/Implementation||how was it built? how does the final implementation relate to the theory you set forth in chapter three?|
|5) Evaluation||unfortunately, research must have questions, and questions must be evaluated. else, you end up saying something like, "my system behaves exactly as it behaves." How well does your work acheive the claims you set forth in the introduction and theory sections? how do you know that it "works"?|
|6) Conclusion||and now, you're done. you can put in future work, but i haven't a clue why you would; it's likely that you'll never pursue such work. you'd be better off just summarizing your main argument, how the design of the technology supports the argument, and how the evaluation supports your hypotheses.|