README.md

FedNLP

Git repo for paper "Federated Learning with Convex Global and Local Constraints"

Overview

task	datasets
Linear Equality Constrainted Quadratic Programming (LCQP)	synthetic data (automatically generated when running the scripts)
Nayman-Pearson Classification (NP)	breast-cancer-wisc; adult-a, monks-1 download 🔗
Classification with Fairness Constraints (Fairness)	breast-cancer-wisc; adult-a, monks-1 download 🔗

Installation

git clone https://github.com/PL97/Constr_FL.git

Datasets

Download the dataset using the link in the table. Create a new folder named $data/$ at root and put the downloaded data into it.

Usage

Linear Equality Constrainted Quadratic Programming (LCQP)

cd LCQP

python lcqp.py

Nayman-Pearson Classification (NP)

centralized training 👇

cd NP/decentralized_alg/

python np.py --dataset [choose from "breast-cancer-wisc", "monks-1", "adult"] \
  --n_clinet [e.g., 1, 5, 10, 20]

Classification with Fairness Constraints (Fairness)

centralized training 👇

cd NP/Fairness/wglobal/

python fairClassification.py --n_client [e.g., 1, 5, 10, 20] --repeat_idx [random seed, e.g., 0, 1, 2]

How to cite this work

@article{
he2024federated,
title={Federated Learning with Convex Global and Local Constraints},
author={Chuan He and Le Peng and Ju Sun},
journal={Transactions on Machine Learning Research},
issn={2835-8856},
year={2024},
url={https://openreview.net/forum?id=qItxVbWyfe},
note={}
}