synthetic_data

synthetic_data is a Python module that enables generation of synthetic data from real data. The module enables generation of data which can be distributed easily without revealing private information.

Dependencies

• Python (>= 3.6)
• NumPy (1.17.0)
• Pandas
• SciPy
• scikit-learn
• tensorflow (1.13.1)
• progress
• psutil
• tqdm
• matplotlib
• seaborn

Installation

1. Download the repository to your machine using the command below which will generate a folder synthetic_data

git clone https://github.com/TheRensselaerIDEA/synthetic_data.git

2. Go inside the folder synthetic_data

cd synthetic_data

3. Install all dependencies using the command below

python3 setup.py install

Usage

The step by step guide to use the package and all its functionalities is available as a Jupyter notebook here

GPU Support

The package also supports the use of GPUs to facilitate the training of the model. To access GPUs, you need to install tensorflow-gpu==1.13.1 and then define the visible CUDA devices.

Install tensorflow-gpu

Note that the package is developed on Tensorflow 1.13.1 and thus, it is recommended to use the same version for tensorflow-gpu to avoid incompatibility issues.

pip install tensorflow-gpu==1.13.1

Set visible devices

Place this code at the start of your Python script. Considering the visible CUDA devices are 2 and 3, the code shall be:

import os
import tensorflow as tf

os.environ["CUDA_VISIBLE_DEVICES"] = "2, 3"