Benchmark

class pyapprox.benchmarks.Benchmark

Bases: OptimizeResult

Contains functions and results needed to implement known benchmarks.

A benchmark can be created with any attribute. Only fun and variable are required. Below are these two required attributes and other optional attributes used in different PyApprox Benchmarks

Notes

Use the keys() method to see a list of the available attributes for a specific benchmark

Attributes:

funcallable

The function being analyzed

variableJointVariable

Class containing information about each of the nvars inputs to fun

jaccallable

The jacobian of fun. (optional)

hesscallable

The Hessian of fun. (optional)

hesspcallable

Function implementing the hessian of fun multiplied by a vector. (optional)

mean: np.ndarray (nvars)

The mean of the function with respect to the PDF of var

variance: np.ndarray (nvars)

The variance of the function with respect to the PDF of var

main_effectsnp.ndarray (nvars)

The variance based main effect sensitivity indices

total_effectsnp.ndarray (nvars)

The variance based total effect sensitivity indices

sobol_indicesnp.ndarray

The variance based Sobol sensitivity indices