Sampling¶

Prandtl provides three sampling methods for exploring parameter spaces.

Methods¶

Latin Hypercube Sampling (LHS) — default¶

Space-filling design. Maximizes coverage with minimal points. Best for surrogate modeling.

X, Y = pr.sample(
    pr.analytical.cl_flat_plate,
    bounds=[(-5, 15), (0.01, 0.1)],
    n=100,
    method="lhs",
    seed=42
)

Uniform Random¶

Simple, unbiased. Useful for validation sets.

X, Y = pr.sample(
    func, bounds=[(0, 1), (-2, 2)],
    n=100, method="uniform", seed=42
)

Sobol Sequence¶

Low-discrepancy quasi-random. Deterministic — same input always gives same points.

X, Y = pr.sample(
    func, bounds=[(0, 1), (-2, 2)],
    n=128, method="sobol"
)

Requires n to be a power of 2 for best uniformity.

Using existing data¶

You don't need sample() if you already have data:

surrogate = pr.Surrogate(params=["alpha", "mach"], outputs=["CL"], method="gp")
surrogate.fit(X, Y)  # X: (n_points, n_params), Y: (n_points, n_outputs)

API¶

Parameter	Type	Description
`func`	callable	Takes dict of param values → returns dict of output values
`bounds`	list of (min, max)	One tuple per parameter
`n`	int	Number of samples
`method`	str	`"lhs"`, `"uniform"`, or `"sobol"`
`seed`	int	Random seed for reproducibility