This example shows how to use PiShield for synthesizing realistic tabular data, which is compliant with a set of requirements written as linear inequalities.

Task: training a deep generative model (DGM) for tabular data generation.

Training Data: WiDS dataset, which is used to predict if a patient is diagnosed with a particular type of diabetes named Diabetes Mellitus, using data from the first 24 hours of intensive care. For training the models here, 22K samples were used, each sample having 109 features.

Requirements: 31 linear inequalities capturing relations between the data features.

Example requirement: the value associated with the "maximum level of hemoglobin recorded" column to be greater than or equal to the one associated with the "minimum level of hemoglobin recorded" column.