implementation of back propagation through a computational graph for addition, multiplication, division, subtraction, exponents, and various activation functions.
implementation of a simple neural network node, and a multi layer perceptron with customizable parameters
demo of my neural network on the make moons dataset. the model takes in 3 parameters and outputs 1 (-1/1) for classification. The model is trained using gradient descent using the grad.py implementation and a varying learning rate for Stochastic Gradient Descent. The model is able to achieve an accuracy of 100% after 100 steps on the training set, and is also able to achieve an accuracy of 100% on never seen before testing data.