This project is an in-depth exploration of Deep Neural Networks (DNNs). By clearly defining the components of a DNN, and how they function, a much deeper understanding of these models can be cultivated. The full paper dives into the implementation in great detail, covering everything from definitions to proofs. I hope this project can serve as a valuable reference for those looking to understand DNNs in a technical manner. This project was inspired by Omar Aflak's Implementation, and much of the functional code/structure was based on their repository.

If you wish to learn more about DNNs, start by reading the included paper to gain a better understanding.

Implement your own models by creating an array of layers, note the final layer shape should be the number of features you wish to output. In this case, only 1 neuron output indicates a binary classifier.

network = [
    Dense(2, 3),
    tanh(),
    Dense(3, 1),
    tanh()
]

Train the model using the following function, defined in the network.py file:

train(network, mse, mse_prime, X, Y, epochs=10000, learning_rate=0.1)