gp_api.utils package

Submodules

gp_api.utils.fit_compact module

Utilities for the compact kernel

gp_api.utils.fit_compact.fit_compact_nd(x_train, y_train, whitenoise=0.0, sparse=True, xpy=None, order=1, **kwargs)

Fit data for an arbitrary function using the compact kernel

Parameters

  • x_train (array like, shape = (npts, dim)) – Training data samples

  • y_train (array like, shape = (npts,)) – Training data values

  • whitenoise (float, optional) – White noise kernel threshold

  • sparse (bool, optional) – Use sparse matrix operations from scikit-sparse?

  • xpy (ModuleType, optional) – A module, either numpy or cupy. cupy is not yet supported.

gp_api.utils.fit_compact.train_function(true_pdf, dim, limits=None, train_res=10, xnoise=0.0, ynoise=0.0, seed=10)

Generate training data for an arbitrary function on a grid

Parameters

  • true_pdf (function) – A function to evaluate the pdf on

  • dim (int) – Number of dimensions

  • limits (array like, shape = (dim, 2)) – List of (min, max) pairs for each dimension

  • res (float) – Sample resolution for training

  • xnoise (float) – Input noise for data

  • ynoise (float) – Output noise for data

  • seed (np.random.RandomState or int, optional) – Seed for the random number generator

gp_api.utils.fit_matern module

Utilities for the Matern kernel

gp_api.utils.fit_matern.fit_matern_nd(x_train, y_train, whitenoise=0.0, nu=0.5, xpy=None, **kwargs)

Fit data for an arbitrary function using the compact kernel

Parameters

  • x_train (array like, shape = (npts, dim)) – Training data samples

  • y_train (array like, shape = (npts,)) – Training data values

  • whitenoise (float, optional) – White noise kernel threshold

  • sparse (bool, optional) – Use sparse matrix operations from scikit-sparse?

  • xpy (ModuleType, optional) – A module, either numpy or cupy. cupy is not yet supported.

gp_api.utils.hypercube module

Utilities for hypercubes

gp_api.utils.hypercube.sample_hypercube(limits, res)

Generate a list of points for a hypercube with shape (res**dim, dim)

Parameters

  • limits (array like, shape = (dim, 2)) – List of [min,max] pairs for each dimension

  • res (int) – Sample resolution in one dimension

gp_api.utils.multigauss module

Utility for multi-dimensional Gaussians

class gp_api.utils.multigauss.Multigauss(ngauss, dim, limits=None, std_offset=0.1, seed=10)

Bases: object

Multi-Gaussian probability density

This is a useful distribution for testing KDE accuracy.

rvs(n_sample)

Draw random samples

Parameters

n_sample (int) – Number of random samples to draw

sample_grid_gp(sample_res=10)

Sample a grid of hypercube points with GP approximate y values

train(x_train, y_train, **kwargs)

Train a compact GP approximate using training data

Parameters

  • x_train (array like, shape = (npts,)) – Training samples matching the pdf

  • y_train (array like, shape = (npts,)) – Training values matching the pdf

train_matern(x_train, y_train, **kwargs)

Train a Matern GP approximate using training data

Parameters

  • x_train (array like, shape = (npts,)) – Training samples matching the pdf

  • y_train (array like, shape = (npts,)) – Training values matching the pdf

training_grid(train_res=10, xnoise=0.0, ynoise=0.0)

Generate a training grid

Parameters

  • train_res (int, optional) – Number of training points on side of hypercube

  • xnoise (float, optional) – Scaling on noise added to x_train

  • ynoise (float) – Scaling on noise added to y_train

gp_api.utils.utils module

General utilities used by the package

gp_api.utils.utils.fit_compact_nd(x_train, y_train, whitenoise=0.0, sparse=True, xpy=None, **kwargs)

Fit data for an arbitrary function using the compact kernel

Parameters

  • ndarray[npts (x_train) –

  • dim] (Training data samples) –

  • ndarray[npts] (y_train) –

  • whitenoise (Whitenoise kernel threshold) –

  • sparse (Use sparse matrix operations from scikit-sparse?) –

  • xpy (numpy equivalent (alternatively cupy)) –

gp_api.utils.utils.fit_multigauss(n_gauss, dim, limits=None, seed=10)

Fit a number of Multivariate Normal distributions in n dimensinos

Inputs:

n_gauss: number of gaussians to generate dim: number of dimensions limits ndarray[:,2]: list of [min,max] pairs seed: numpy random state or int

gp_api.utils.utils.multigauss_nd(n_gauss, dim, limits=None, train_res=10, sample_res=10, xnoise=0, ynoise=0, whitenoise=0.0, seed=20211222, sparse=True, xpy=None)

Model an n dimensional gaussian and provide a sample cube

Inputs:

n_gauss: number of gaussians to generate dim: number of dimensions limits ndarray[:,2]: list of [min,max] pairs train_res: sample resolution for hypercube training data sample_res: sample resolution for hypercube samples xnoise: input noise for data ynoise: output noise for data whitenoise: noise for whitenoise kernel seed: numpy random state or int sparse: Use sparse matrix operations from scikit-sparse? xpy: numpy equivalent (alternatively cupy)

gp_api.utils.utils.multigauss_samples(n_gauss, dim, n_sample, limits=None, seed=20211222)

Model an n dimensional gaussian and draw random samples

Inputs:

n_gauss: number of gaussians to generate dim: number of dimensions n_sample: number of random samples to draw limits ndarray[:,2]: list of [min,max] pairs seed: numpy random state or int

gp_api.utils.utils.sample_list_nd(limits, res)

Generate a list of points for a hypercube with shape (res**dim, dim)

Inputs:

limits ndarray[:,2]: list of [min,max] pairs res int: sample resolution in one dimension

gp_api.utils.utils.train_function(true_pdf, dim, limits=None, train_res=10, xnoise=0.0, ynoise=0.0, seed=10)

Generate training data for an arbitrary function on a grid

Inputs:

true_pdf: a function to evaluate the pdf on dim: number of dimensions limits ndarray[:,2]: list of [min,max] pairs res: sample resolution for TRAINING xnoise = input noise for data ynoise: output noise for data seed: numpy random state or int

Module contents