Setup to predict a final course grade; the course structure is 4 exams and 20 homework assignments. This network performs gradient descent on 12 input parameters, which are the first 2 exam grades and first 10 homework grades of the course. The neural network uses 7 neurons in 1 hidden layer, Sigmoid as the activation function in each neuron, and a learning rate of 0.01.
Why bother trying machine learning in C++ instead of Python? When learning this stuff for the first time Python can obfuscate the math behind things by hiding it in fancy pre-built libraries with near-English sounding names; I believed it would be more mathematically helpful to use a lower level language to learn the methods a little better, alongside using Python of course.