Source code of ICDAR 2025 submitted paper 'Hypergraph-Driven Tabular Data Synthesis with Multi-Objective Optimization'
2025.3
HyperTab, a novel generative model based on hypergraph modeling and adversarial training. HyperTab leverages hypergraphs to capture the permutation invariance inherent in tabular data. By utilizing a hypergraph encoder, it explicitly models the interaction relationships between features, ensuring that the synthesized data accurately reflects global dependencies.
Create the main environment with requirements.txt. This environment will be used for training model and sampling synthetic data
Python version: 3.10
Create environment
conda create -n hypertab python=3.10
conda activate hypertab
pip install -r requirements.txt
Create another environment with synthcity.txt. This environment will be used for evaluation.
conda create -n synthcity python=3.10
conda activate synthcity
pip install -r synthcity.txt
Download raw dataset:
python download_dataset.py
Process dataset:
python process_dataset.py
For different data sets, run different train files to train the model.
python train_xxx.py
After training the model, run different generate files for different data sets to generate tabular data.
python generate_xxx.py
To evaluate the quality of the synthesized tabular data, we employed several metrics.
Density estimation of single column and pair-wise correlation (link)
python eval/eval_density.py --dataname [NAME_OF_DATASET] --model [METHOD_NAME]
Machine Learning Efficiency(MLE)
python eval/eval_mle.py --dataname [NAME_OF_DATASET] --model [METHOD_NAME]
Distance to Closest Record (DCR)
python eval/eval_dcr.py --dataname [NAME_OF_DATASET] --model [METHOD_NAME]
python eval/eval_stcp.py --dataname [NAME_OF_DATASET] --model [METHOD_NAME]
We appreciate your citations if you find this repository useful to your research!