Adaptive Stratification

This package provides an implementation of the adaptive stratification sampling method to estimate quantities of interest of the form Q = E(f(Y)), where the random vector Y follows a d-dimensional uniform distribution on the unit-cube and f is a given function.

Example: Using the Sampler

# Import the module containing the sampling routines
from stratification import AdaptiveStratification

# Create a sampler for function func
sampler = AdaptiveStratification(func, d, N_max, N_new_per_stratum, alpha, type='hyperrect')

# Solve (return a tuple)
result = sampler.solve()

Input arguments:

• func: implementation of given function of interest that defines the quantity of interest. It needs to be callable, accepting one m-times-n-dimensional numpy array as input and returns a m-dimensional numpy array;
• d: dimension of the stochastic domain;
• N_max: number of total samples to be used;
• N_new_per_stratum: targeted average number of samples per stratum, controlling the adaptation;
• alpha: number between zero and one, defining the hybrid allocation rule;
• type: type of tessellation procedure, i.e., via hyper-rectangles (type='hyperrect') or simplices (type='simplex')

More Information

See the Github repository for more details.