GPC

GPC(dataset[, outputs, seed])

Methods

GPC.build_model([seed, continuous_kernel, ...])

Compile a marginalized pymc model for the GP.

GPC.draw_point_samples(points, *args[, ...])

Draw posterior samples at supplied points

Attributes

class gumbi.regression.GPC(dataset: DataSet, outputs=None, seed=2021)

Bases: GP

build_model(seed=None, continuous_kernel='ExpQuad', period=None, heteroskedastic_inputs=False, heteroskedastic_outputs=True, sparse=False, n_u=100, ARD=True)

Compile a marginalized pymc model for the GP.

Each dimension in continuous_dims is combined in an ExpQuad kernel with a principled \(\text{InverseGamma}\) prior for each lengthscale (as suggested by Michael Betancourt) and a \(\text{Gamma}\left(2, 1\right)\) prior for variance.

Parameters:

  • seed (int, optional.) – Random seed. If None, seed is used.

  • continuous_kernel ({'ExpQuad', 'Matern32', 'Matern52', 'Exponential', 'Cosine', or 'Periodic'}) – Covariance function to use for continuous dimensions. See pymc docs for more details.

  • period (ParameterArray, optional) – A single parray of length 1 with one layer for each continuous_dims by name containing the period of the kernel, if periodic-like kernel is used.

  • heteroskedastic_inputs (bool, default False) – Whether to allow heteroskedasticity along continuous dimensions (input-dependent noise).

  • heteroskedastic_outputs (bool, default True) – Whether to allow heteroskedasticity between multiple outputs (output-dependent noise). Not yet implemented.

  • sparse (bool, default False) – Whether to use a sparse approximation to the GP.

  • n_u (int, default 100) – Number of inducing points to use for the sparse approximation, if required.

  • ARD (bool, default True) – Whether to use “Automatic Relevance Determination” in the continuous kernel. If _True_, each continuous dimension receives its own lengthscale; otherwise a single lengthscale is used for all continuous dimensions.

Returns:

self

Return type:

GP

draw_point_samples(points, *args, source=None, output=None, var_name='posterior_samples', additive_level='total', increment_var=True, **kwargs)

Draw posterior samples at supplied points

Parameters:

  • points (ParameterArray) – 1-D ParameterArray vector of coordinates for prediction, must have one layer per self.dims

  • output (str or list of str, optional) – Variable for which to make predictions

  • source ({None, dict, az.data.inference_data.InferenceData}) – GP parameters for which to draw samples. Should be the result of find_MAP(), sample(), or _None_.

  • var_name (str, default "posterior_samples") – Name to assign new variable to contain conditional predictions.

  • additive_level (str, default "total") – Level of additive GP at which to make predictions.

  • increment_var (bool, default True) – Whether to append ‘_’ to the end of _var_name_ if it already exists in model.

Returns:

samples – Samples as a ‘Parray’

Return type:

parray