Geodesic Kernel flow for Semi-supervised learning on a Mixed-variable Tabular dataset

This is the origin Pytorch implementation of Geodesic Kernel flow for Semi-supervised learning on a Mixed-variable Tabular dataset.

Warning This code is not cleand. We will be update the code soon.

News(Jun, 01, 2024): The code has been released..

Get Started

To install all dependencies:

pip install -r requirements.txt

Dataset

You can access the well pre-processed datasets from Google Drive, then place the downloaded contents under ./all_dataset

Quick Demo

0. If you want to run the Semi setting, please path to cd GKP_Semi.
1. Download datasets and place them under ./all_dataset
2. We provide all experiment scripts for demonstration purpose under the folder ./runfile. For example, you can evaluate on churn and adult datasets by:

bash ./runfile/churn.sh 

bash ./runfile/adult.sh 

3. If you want to experiment with all datasets, run the bash file from run_a to run_d.

bash ./runfile/run_a.sh
bash ./runfile/run_b.sh
bash ./runfile/run_c.sh
bash ./runfile/run_d.sh 

Detailed usage

Please refer to run.py for the detailed description of each hyperparameter. We also provide a model where the linear projection of the VSN layer is replaced by KAN. We'll update with more details as they become available.

Citation

If you find this repo useful, please cite our paper.

Will be update