Inprocessing-debiasing-methodology-for-multiple-sensitive-attributes

Description

This project implements an adversarial framework to debias deep learning models for multiple sensitive attributes/confounding attributes using gradient reversal and a dynamic loss balancing technique. The approach achieves a fairer model that shows less disparity amongst minority subgroups.

Features

• Adversarial debiasing framework
• Gradient reversal technique
• Dynamic loss balancing
• Support for multiple sensitive attributes
• Improved fairness for minority subgroups

Installation

This project uses a requirements.txt file for package management and a setup.py file for installation. To set up the project:

1. Clone the repository: git clone https://github.com/JVedant/Inprocessing-debiasing-methodology-for-multiple-sensitive-attributes.git
2. cd Inprocessing-debiasing-methodology-for-multiple-sensitive-attributes
3. Run the setup script: python setup.py install

Usage

To get started with the project:

1. Navigate to the source directory: cd Inprocessing-debiasing/src

2. Run the main script to train the model: python main.py

The project structure includes the following key files:

• config.py: Configuration settings
• dataset.py: Dataset handling
• inference.py: Inference functions
• losses.py: Loss functions
• main.py: Main script to run the training process
• models/debias_models.py: Debiasing model implementations
• training.py: Training functions
• validation.py: Validation functions

Contributing

Contributions to improve the project are welcome. Please feel free to submit a Pull Request.

Contact

[Vedant Joshi] [vedantjoshi@asu.edu]