Dissertation code and report.
July 11th, 2016 - 1 minute read
I know I should have been more active on github. I could have been but when working on your own it just feels easier? (hard mode without version control).
There is an awful lot that I would have done to improve it, I really really wanted to squeeze deep learning in there but learning all of that from scratch was just a tiny bit much.
The dissertation is about the self-organising map, k-means and using two distance measures. There are some obvious flaws with the project to start with such as the images not being invariant to scale or rotation (my supervisor sprung this on me in the middle of my viva which was a huge problem). Perhaps I'll make scale and rotation invariant later when I add the adjusted rand index.
My viva (the interview) got 80/100.
The code and report also got 80/100.
My supervisor didn't read a draft.
I got an upper second class degree overall, a couple percent away from a first...