CopulaSynthGenerator<T>

Models & Types · AiDotNet.NeuralNetworks.SyntheticData

Copula-Based Synthesis generator that models marginal distributions independently and couples them with a Gaussian copula to capture inter-feature dependencies.

Inheritance: SyntheticTabularGeneratorBase<T> → CopulaSynthGenerator<T>

For Beginners

This method works like building synthetic data from two ingredients:

Ingredient 1 — Marginal shapes: Each feature’s histogram is learned independently (e.g., “Age is bell-shaped around 40”).

Ingredient 2 — Correlation structure: How features move together (e.g., “Age and Income go up together”).

To generate: sample correlated random values, then map each one to its feature’s histogram. The result preserves both individual distributions and pairwise correlations.

How It Works

The generator operates in three phases:

  1. Fit marginal distributions for each feature using the empirical CDF (the sorted observed values)
  2. Transform data to uniform [0,1] via marginal CDFs, then to standard normal via inverse CDF
  3. Fit a Gaussian copula (correlation matrix) on the normal-transformed data

To generate:

  1. Sample from multivariate normal using the learned correlation matrix
  2. Transform back to uniform via standard normal CDF
  3. Transform to original scale via inverse marginal CDFs (quantile functions)

Constructors

Constructor Summary
CopulaSynthGenerator(CopulaSynthOptions<T> options) Initializes a new instance of the CopulaSynthGenerator class.