INTRODUCTION

Motivation

This is a personal project to understand more deeply how neural networks work and maybe later test with edge cases like self modifying networks.

Achieved goals

I've achieved to make a model to predict the types of flowers in the Iris Dataset as you can see in the following graph.

I've also achieved to correctly identify the numbers in the images of the Mnist Dataset with 28x28 images that look like this. Each model stored as a pickle obj in the models folder achieved the accuracy in the final test that is shown in the name. Here is the graph of the learning of one of them. Theres is also a interactive teste of model in which you draw numbers in a pygame window and when you press enter it tries to guess the number drawn (It isnt very good for certain numbers) the test is in the file mnist_model_tester.py follow the installation instrucctions to use it

Installation and usage

To use the Neural network class

pip install -r requirements.txt

To use the mnist model teste

pip install -r requirements_mnist_model_tester.txt

To train a model for the mnist dataset

pip install -r requirements_mnist_model_train.txt

Then run the python file that you want to test