There are two versions of the neural network here:
- An old one (all .py files): I made it almost a year ago (uploaded later tho); back then I had no much experience in machine learning and just finished my first year at the university.
- A new one (.ipynb files): Was made as a university assignment with more experience in the field.
New NN has better stability, flexibility, and is producing good results (unlike the old neural network). The main differences between those two are:
- The code structure;
- Weights initialization: the new one uses numpy's normal distribution;
- New NN uses delta rule in order to achieve better performance;
P.S. If you, by some chance, want to see my "thinking process" during the neural network implementation, here's the mind-mapping whiteboard. It's messy and quite hard to understand, but can be useful anyways (and sorry for the bad quality) :)
