First neural network in the Deep Learning Nanodegree course by Udacity. This was a relatively light introduction to the world of model architecture, but it was good practice getting some of the low-level math down in the my_answers.py file.
The most difficult part of the project for me was figuring out the dimensions of the arrays and getting them into the correct format.
Once that was done, I had a difficult time understanding exactly what format the loss needed to be in when passed back into the main function.
In order to train this project for yourself, download the data set zip and put it into the same folder as the main project and the my_answers.py files.
Once I had the calculations figured out, I spent a lot of time playing around with the hyperparameters on this and managed to get to a pretty low loss score. The final project that I handed in was not the lowest but was more than sufficient to pass.
This project was not my first neural network, but it was one of the more furn forays into the field. I finished hoping that more advanced networks would be this cool.
Check out the pdfs to see my feedback and what the notebook looked like when I ran it after the fact.