This Repo is a learning experience, and we will be covering the majority of how to build neural networks
-
The Neural network will be a feed-forward model with a multi-layer architecture.
-
This network will be trained on MNIST, fashion_mnist, mnist_corrupt, and mnist_corrupt.
-
After implementation is finished, experiments shall commence.
In hopes of making this project, only a few exceptions shall be made:
-
For data Keras will be used to only load the data
-
Numpy will be used in this project for most of the project, mainly to get a close result or at least accurate results
-
Due to some technical difficulties, some files and ideas may have been brought here though in this version, it has been resolved, and the project went through redesigns.
# Changed Plans:
Currently, the project is going through some redesigns since some Factors didn't go as planned.
In the testing phase, to check if building it was possible, but due to some difficulties,
some base build redesigns will take place.
The pipeline exists, but will take time to be made since a web view is also planned to be implemented.
The project will stay true to itself where it will be built from scratch in python (only using numpy).
The frontend is planned to be built in HTML.
**Note: This project has split in two(this one the neural network, and machine learning from scratch)**