How do I talk about methodology mistakes In a scientific project report?

Question

I'm writing up my 3rd year engineering project (The equivalent of a dissertation in other subjects), and I'm struggling to write the chapter in which my method failed I couldn't get any results.

My project is on Neural Networks, and when I started my project I tried to program my own Neural Network from scratch, however this did not work properly (The feed forward part worked, but because of the convoluted way I had programmed it I couldn't implement the back propagation part). After I realized this I used an off-the-shelf program to create my neural network, which worked much better.

The problem is that I must have spent about a quarter of the year programming my own Neural Network from scratch, so I want to include it as a chapter in my report to show the marker that I have put over the recommended 300 hours work into my project, and that I understand the inner working of Neural Networks.

How should I write this chapter? So far I have started the chapter as if it is going to work, then I wrote about exactly what I have done and how it works, and then at the end where the results should be I want to explain that the custom implementation of the neural network meant that I couldn't implement back propagation, and therefore couldn't get results. But I can't do this without it looking like I have lazily given up on that chapter, and abruptly given an excuse to move on to the next chapter (The working version of the neural network).

Answer 1

You wrote:

I want to include it as a chapter in my report to show the marker that I have put over the recommended 300 hours work into my project

Nooooooooo.

Don't do that. This is a postgraduate degree, and you're expected to take the research seriously. It's not a 9-5 grunt job or prison sentence where you turn up, serve your time, and get the reward at the end of your allotted time.

Show what you've learnt. That's what the markers are interested in. You say that you've learnt the inner workings of neural networks. That's great. So demonstrate and document what you've learnt.

And don't document the tortuous process you used. Don't describe how you went from A to B via C,D,E,...X,Y.

Just report that you started at A, ended at B, you've found that the shortest route between those points is {whatever it is}, and briefly document the dead ends and traps that future researchers should avoid.

Answer 2

to show the marker that I have put over the recommended 300 hours work into my project

So what? It's a thesis, not prison time. Your task in your thesis (or project) is to solve a given problem, not serve a given amount of time. To be brutally honest, the time that you spent on what was presumably a rather bad idea in the first place (re-implementing something for which apparently suitable standard software exists) does not make your thesis better in any way. Detailing your failed labouring will not make your project seem better. Rather the contrary.

To answer your title question:

How do I talk about methodology mistakes In a scientific project report?

You don't, unless:

• The methodology mistakes are somehow common in your field, i.e., you would assume if another researcher would start the same project, he would likely make the same mistake (in that case there is a lesson learned to take away from the failed attempt, i.e., that the standard approach does not work in this case).
• The methodology mistakes have, inadvertently, led to a different interesting result / observation that you did not expect.

Arguably, both is not true in your case. As such, I would consider this quarter of a year as sunk costs in terms of thesis time.

Answer 3

I'm afraid I disagree with the comments above.

The purpose of a 3rd-year project is not simply to solve a problem - it is to train the student. Learning from doing and learning from making mistakes is an integral element of this. For exactly this reason, the report is not just a scientific report of results, but also a report on the process, and I would say that there is no problem in reflecting some of the 'tortuous path' in the report.

Of course, it would be wise not to dwell too long on the mistakes. I would briefly describe how the 'wrong' path started, how you found out that it was wrong, and - especially - describe how you now understand why it is wrong, and why the later method is better. If this is done briefly and honestly, I would very much appreciate reading it, and it would score points with me as a reviewer.