LinearDGP

LinearDGP generates datasets where the target is a linear combination of input features. The complexity parameter controls the coefficient structure: how many features carry signal and how uniformly their coefficients are distributed.

Quick Start

import synthbench
from synthbench import BenchPipeline, LinearDGP

dgp = LinearDGP(complexity="medium", task_type="regression", random_state=0)
pipeline = BenchPipeline(dgp)
result = pipeline.run(n_samples=500, n_features=10, random_state=42)

print(result.X.shape)   # (500, 10)
print(result.y.shape)   # (500,)
print(list(result.metadata.keys()))

# Signal importances sum to 1.0
importances = result.metadata["signal_feature_importances"]
print(sum(importances.values()))  # 1.0

Parameters

Parameter	Default	Description
complexity	"medium"	Controls coefficient structure: "low" uses fewer features, "high" uses more with varied weights
task_type	"regression"	"regression" for continuous target, "classification" for binary labels
random_state	0	Integer seed for reproducibility
class_weight	0.5	(Classification only) Fraction of samples in the positive class

Notes

• Low complexity: Few informative features with near-equal coefficients.
• Medium complexity: Moderate number of informative features.
• High complexity: Many features with varying coefficient magnitudes.
• Signal feature importances reflect the squared coefficient magnitudes, normalized to sum to 1.0.
• Noise features receive exactly 0.0 importance in signal_feature_importances.