ML From Scratch

This repo contains my implementations of machine learning algorithms from scratch using numpy.

The idea behind this project is to gain a deeper understanding to different ML algorithms, tricks, and practices, by implementing them from scratch and seeing their impacts. The project tries to approach ML from a very first-principles view, trying to uncover the core ideas hidden behind the abstractions provided by many libraries.

Gradient Checking

Every layer has a logic for forward pass and backward pass. The best way to check the correctness of the implementation is to do a gradient check. The gradients for a layer can be checked using following command:

python grad_check_<>.py --function <function name>

Implementations

The following mini projects are implemented:

1. Classification on MNIST data using a dense network.

Datasets Used

You can find the datasets used in the scripts from the following links:

• MNIST