a fully-functional neural network implemented from scratch in C++ using the Eigen Matrix library. most neural networks are implemented in python and their really aren't much neural networks implemented from scratch in c++. i chose to use c++ as despite it being harder to program due to limited information, the network would perform much faster and would be more efficient.
- neurons
- layers
- activation
- loss
- accuracy
- derrivatives
- backpropagation
- optimization
when building the network, we will be using an insignificant dataset, the spiral dataset, to test our neural network. when the network finishes completion, we will use this hand-written digit recognition dataset: https://www.kaggle.com/competitions/digit-recognizer/data.
a small visualiser is created with the olcPixeGameEngine to paint digits and test out the network.