The Synthetic Data Generator Evaluative Visualization Tool uses Generative Adversarial Networks (GANs) to create and evaluate synthetic data, crucial for privacy preservation and data balance in machine learning. This tool provides real-time visualization and metrics to give instant feedback on synthetic data quality.
Key features:
- Iterative, user-guided generation to minimize computational load, especially for large datasets.
- Customizable parameters such as learning rate, batch size, and more, to fine-tune the generation process.
- Real-time metrics like losses and KL divergences displayed on the left, allowing you to assess data quality at each iteration.
- Dynamic visualizations that show the proximity of synthetic data to original data, helping users decide when to stop the generation process.
-
Clone the repository:
git clone https://github.com/anonymsynthetic/SDGEV.gitcd ./Synthetic-Data-Generator-Evaluative-Visualization-Tool -
Create and activate a virtual environment:
python3 -m venv sdenvsource sdenv/bin/activate -
Install the required packages:
pip install -r requirements.txt -
Run the app:
./sdenv/bin/python -m streamlit run src/main.pyAfter running, open the provided local URL (e.g., http://localhost:8501) in your browser.
Key features:
- Parameter selection: Choose parameters like learning rate, batch size, etc., before starting the synthetic data generation.
- Real-time feedback: Once generation starts, the app displays performance metrics such as losses and KL divergences on the left panel.
- Visualization: The tool visualizes the similarity between synthetic and original data in real-time, helping you evaluate data quality and stop the process when desired.
For more details on the project, read the full research paper here: https://github.com/Ahmetyasin/Synthetic-Data-Generator-Evaluative-Visualization-Tool/blob/main/docs/Paper.pdf
