Neural Networks Library from Scratch

the goal in this project is to create a small and modular library to implements a classic neural networks. I was insipred by the architecure of pytorch and tensorflow to create my classes and functions.

I'm doing this project because I'm still using libraries like pytorch and tensorflow and I wanted to implement a neural network myself from scratch.

Features

Layers

• ✅ Linear Layer

Activation Functions

• ✅ ReLU
• ✅ Sigmoid
• ✅ Tanh
• ✅ Softmax

Loss Functions

• ✅ MSE
• ✅ CrossEntropy
• ✅ BinaryCrossEntropy

Optimizers

• ✅ SGD
• ✅ Adam

Example of classification (see more in the example directory)

Multi classification

Multi classification on the mnist dataset

Binary classification

Example of regression

Linear regression