XAI Exercise

This is a repository for the XAI (Explainable AI) exercise.

Description

The exercise consists of training a machine learning model to classify handwritten digits from the MNIST dataset, and then creating a web application to visualize and explain the model's predictions using SHAP (SHapley Additive exPlanations).

Files

The repository contains the following files:

cfel.py: Python class containing the code to compute the counterfactuals of the model.
LSM.py: Python script containing the code for the LocalSurrogateModels.
process.py: sample to encode various types of data.
vizualization.py: script that provide the shap values.
main.py: script that contains a sample of using entire flow.
data.csv: generated data for testing purposes.

Requirements

To run the code, you need to have the following libraries installed:

pandas
numpy
matplotlib
shap
scikit-learn

Usage

Create a virtual environment

python3 -m venv venv

Activate the virtual environment

source venv/bin/activate

Install the required packages

pip install -r Requirements.txt

Run the code

python main.py

Deactivate the virtual environment

deactivate

danciugaby/XAI