Multi-Layer Perceptron Project

This repository hosts a project for the implementation of a multi-layer perceptron. The project aims to preprocess data, train a neural network on it, and finally evaluate the performance of the trained model.

Project Structure

The repository is organized as follows:

1. RemovingMissingDatas.py: This script is for removing any missing data from the dataset.
2. LettersToNumbersConversion.py: This script is responsible for converting categorical values (letters) into numerical values.
3. Normalization.py: This script normalizes the converted numerical data.
4. Main.py: This is the main script where the multi-layer perceptron model is defined, trained and evaluated.

Setting Up and Running

Clone the repository: Start by cloning this repository to your local machine using git clone.

Install dependencies: You'll need to install several Python packages to be able to run this project. You can do this by running pip install -r requirements.txt.

Preprocessing: Run the removingMissingDatas.py, lettersToNumbersConversion.py, and Normalization.py scripts to preprocess the data. Run the scripts in the specified order.

Edit parameters in main.py

Within the main.py script, you can adjust the following metaparameters:

• max_epoch: The maximum number of training epochs.
• err_goal: The desired error goal to achieve.
• disp_freq: The frequency of display updates.
• lr_vec: The learning rate values.
• K1_vec: The first set of node numbers in the hidden layers.
• K2_vec: The second set of node numbers in the hidden layers.

Train the Model: Once your data is preprocessed and metaparameters are set, you can run the main.py script to train and evaluate your multi-layer perceptron model.

The dataset source

https://archive.ics.uci.edu/dataset/73/mushroom