aquakin.DGSMResult#

class aquakin.DGSMResult(input_names, dgsm, sobol_total_bound, std_error, output_variance, n_samples, n_valid, seed, ranges, output_name=None)[source]#

Bases: object

Result of dgsm().

Variables:

  • input_names (list[str]) – Names of the uncertain inputs, matching the rows of every array.

  • dgsm (jnp.ndarray) – The derivative-based global sensitivity measure nu_j = E[(d output / d z_j)^2], shape (d,).

  • sobol_total_bound (jnp.ndarray) – Upper bound on the Sobol total-order index of each input, S_j^tot <= nu_j (b_j - a_j)^2 / (pi^2 Var(f)) for z_j uniform on [a_j, b_j] (Lamboni, Sobol & Kucherenko 2013). Dimensionless and directly comparable across inputs – the AD-accelerated replacement for a variance-based Sobol total index.

  • std_error (jnp.ndarray) – Monte-Carlo standard error of sobol_total_bound (convergence indicator). Shrinks like 1/sqrt(n_valid).

  • output_variance (float) – Variance of the scalar output over the sample.

  • n_samples (int) – Number of quasi-random points actually drawn (a power of two).

  • n_valid (int) – Number of points with a finite output and gradient (others skipped). For a vector-valued fn this is counted per output – a sample non-finite in another output is not dropped from this one – so different outputs may report different n_valid.

  • seed (int) – Seed of the scrambled-Sobol sampler – fixing it makes the result bit-for-bit reproducible.

  • ranges (jnp.ndarray) – The (d, 2) input ranges used.

Parameters:

  • input_names (list[str])

  • dgsm (Array)

  • sobol_total_bound (Array)

  • std_error (Array)

  • output_variance (float)

  • n_samples (int)

  • n_valid (int)

  • seed (int)

  • ranges (Array)

  • output_name (str | None)

__init__(input_names, dgsm, sobol_total_bound, std_error, output_variance, n_samples, n_valid, seed, ranges, output_name=None)#
Parameters:

  • input_names (list[str])

  • dgsm (Array)

  • sobol_total_bound (Array)

  • std_error (Array)

  • output_variance (float)

  • n_samples (int)

  • n_valid (int)

  • seed (int)

  • ranges (Array)

  • output_name (str | None)

Return type:

None

Methods

__init__(input_names, dgsm, ...[, output_name])

ranked()

Return (name, sobol_total_bound) pairs sorted by decreasing bound.

Attributes

output_name

input_names

dgsm

sobol_total_bound

std_error

output_variance

n_samples

n_valid

seed

ranges

ranked()[source]#

Return (name, sobol_total_bound) pairs sorted by decreasing bound.

Return type:

list[tuple[str, float]]