Dataset fitting

Module that contains the classes for performing interpolation or regression of a specified dataset and then evaluating its function value at any given position

class topsearch.potentials.dataset_fitting.DatasetInterpolation(model_data: ModelData, smoothness: float = 0.0)

Class to compute and evaluate a radial basis function interpolation fitted to a given dataset.

model_data

The dataset which we interpolate

Type:

class

smoothness

Smoothness parameter for the kernel in RBF interpolation

Type:

float

model

The scipy interpolation model that can be fit and queried

Type:

class

function(position: NDArray[Any, Any]) → float

Evaluate the value of the interpolation model

initialise_model() → None

Initialise the interpolation model, a radial basis function interpolation with a thin-plate kernel as implemented in scipy

refit_model() → None

Refit the interpolation model for the current model_data

class topsearch.potentials.dataset_fitting.DatasetRegression(model_data: ModelData, model_rand: int = 1)

Class to fit a regression model to a given dataset using the sklearn multilayer perceptron, and then query its value at any point in space

model_data

The dataset which we regress

Type:

class

model_rand

Random seed passed to MLP fitting

Type:

int

model

The sklearn regression model that can be fit and queried

Type:

class

cv_results

The results of cross-validation fitting of MLP

Type:

dict

function(position: NDArray[Any, Any]) → float

Evaluate the value of the regression model

get_model_error() → float

Return the model error of current fit

initialise_model() → None

Initialise the regression model, a multi-layer perceptron as implemented in sklearn

refit_model() → None

Refit model and recalculate error based on current training