Comparing Synthesizers

PETsARD supports multiple data synthesis methods. You can use different algorithms to synthesize data in the same experiment and compare their results.

Click the below button to run this example in Colab:

---
Loader:
  data:
    filepath: 'benchmark/adult-income.csv'
Preprocessor:
  demo:
    method: 'default'
Synthesizer:
  gaussian-copula:
    method: 'sdv-single_table-gaussiancopula'
  ctgan:
    method: 'sdv-single_table-ctgan'
  tvae:
    method: 'sdv-single_table-tvae'
Postprocessor:
  demo:
    method: 'default'
Evaluator:
  demo-quality:
    method: 'sdmetrics-qualityreport'
Reporter:
  output:
    method: 'save_data'
    source: 'Synthesizer'
  save_report_global:
    method: 'save_report'
    granularity: 'global'
...

Available Synthesizers

Gaussian Copula (sdv-single_table-gaussiancopula)

Synthesizes data by modeling variable dependencies
This is PETsARD’s default method

CTGAN (sdv-single_table-ctgan)

Uses conditional generative adversarial networks (conditional GAN)
Focuses on conditional probability distributions of categorical variables

TVAE (sdv-single_table-tvae)

Uses variational autoencoders for synthesis
Focuses on overall data distribution patterns

By specifying multiple synthesis methods in a single YAML configuration file, you can run multiple algorithms at once and compare their performance using the same evaluation metrics.

Data Preprocessing Custom Synthesis