Principle:Sdv dev SDV Constraint Validation

Knowledge Sources	SDV Documentation SDV
Domains	Data_Quality, Validation
Last Updated	2026-02-14 00:00 GMT

Overview

A validation mechanism that checks whether synthetic data rows satisfy the business rule constraints applied during generation.

Description

Constraint validation enables post-generation verification by checking each row of synthetic data against the defined constraints. It returns a boolean Series indicating which rows are valid, allowing users to measure constraint satisfaction rates and identify violations.

Usage

Use constraint validation after sampling to verify that the generated data satisfies all business rules. This is particularly important for reject-sampling constraints where 100% satisfaction is not guaranteed.

Theoretical Basis

For each constraint, the is_valid method applies the constraint's validation logic to the data:

• Inequality: Check $high\ge low$ (or $high>low$ for strict)
• FixedCombinations: Check that all tuples exist in the training set
• Range: Check $low\le middle\le high$
• ProgrammableConstraint: Run user-defined validation logic

Related Pages

Implemented By

• Implementation:Sdv_dev_SDV_BaseConstraint_Is_Valid