Multilayer Perceptron

The goal of this project is to create a multilayer perceptron neural network from scratch. Which means implementing the mathematical formulas for feedforward, backpropagation & gradient descent (cf mlp_subject.pdf)

Prerequisit:

If you do not have python3, run:
apt-get install python3
To create a virtual environment, run:
python3 -m venv [your_env_name]
Then: source [your_env_name]/bin/activate
Finally:
cd src
pip install --upgrade pip
pip install -r requirements.txt

I. Training

Run python3 train.py to train the model on the dataset that is given by the subject and stored in ../datasets/data.csv.
Use -h to display the usage and the options

II. Comparing

Run python3 compare.py to compare different models (differents optimizers, activation functions, loss functions)

Training Output example:

III. Predicting

Run python3 predict.py to make one prediction on a part of the dataset and calculate our loss.
Use -h to display the usage and the options