This project aimed to write perceptron and neural network classes from scratch, and then use them on classic toy problems such as linear/non-linear function approximation and digit recognition using the MNIST dataset.
The app can be found here: https://immodal.github.io/neural-network/
Figure 1: Handwritten digit recognition
Figure 2: Linear Function Approximator