Neural Network implementation from scratch. I.e., without prebuilt functions other than those contained by basic libraries, e.g. numpy.
This Class is a mathematical exercise for myself, rather than a computational efficiency one. So, the class itself will be usable, and I will be incrementally improving it over time with different features (e.g. Activation funcs., different initialisations, training schemes), but won't ever be as efficient as it could be (at least not yet...)