ariane.lib.gaussian_process_regression¶
- class ariane.lib.gaussian_process_regression.GaussianProcessRegressorBase[source]¶
Base class (interface) for Gaussian process regressors.
- abstract add_data(x_new, y_new, noise)[source]¶
Add new data y_new measured at training locations x_new to the Gaussian process and update the posterior distribution. noise is again the standard deviation of the noise.
Parameters¶
- x_newarray_like
2-D array with new training locations.
- y_newarray_like
1-D array with new training data observed at x_new. The length has to be the same as with x_new.
- abstract fit(x, y, noise)[source]¶
Fit the Gaussian process to data y observed at training locations x with noise standard deviations noise.
Parameters¶
- xarray_like
2-D array with training locations.
- yarray_like
1-D array with training data observed at x. The length has to be the same as with x.
- noisefloat or array_like
Standard deviations of observational noise. If array_like, the length has to be the same as with y.
- abstract property kernel¶
Kernel optimized using training data.
Returns¶
kernel : instance of a kernel class (depending on the underlying GPR implementation)
- abstract property kernel_hyperparameters¶
Hyperparameters of kernel.
Returns¶
kernel_hyperparameters : ndarray of float
- abstract property kernel_length_scales¶
Length scales of kernel.
Returns¶
kernel_length_scales : ndarray of float
- abstract property noise¶
Standard deviations of observational noise.
Returns¶
noise : ndarray of float
- class ariane.lib.gaussian_process_regression.GaussianProcessRegressor(gpr_impl: GaussianProcessRegressorImplementation)[source]¶
Class for wrapping an implementation of a Gaussian process regressor.
- add_data(x_new, y_new, noise)[source]¶
Add new data y_new measured at training locations x_new to the Gaussian process and update the posterior distribution. noise is again the standard deviation of the noise.
Parameters¶
- x_newarray_like
2-D array with new training locations.
- y_newarray_like
1-D array with new training data observed at x_new. The length has to be the same as with x_new.
- fit(x, y, noise)[source]¶
Fit the Gaussian process to data y observed at training locations x with noise standard deviations noise.
Parameters¶
- xarray_like
2-D array with training locations.
- yarray_like
1-D array with training data observed at x. The length has to be the same as with x.
- noisefloat or array_like
Standard deviations of observational noise. If array_like, the length has to be the same as with y.
- property kernel¶
Kernel optimized using training data.
Returns¶
kernel : instance of a kernel class (depending on the underlying GPR implementation)
- property kernel_hyperparameters¶
Hyperparameters of kernel.
Returns¶
kernel_hyperparameters : ndarray of float
- property kernel_length_scales¶
Length scales of kernel.
Returns¶
kernel_length_scales : ndarray of float
- class ariane.lib.gaussian_process_regression.LogGaussianProcessRegressor(gpr_impl: GaussianProcessRegressorImplementation)[source]¶
Class for log-transforming and wrapping an implementation of a Gaussian process regressor.
- add_data(x_new, y_new, noise)[source]¶
Add new data y_new measured at training locations x_new to the Gaussian process and update the posterior distribution. noise is again the standard deviation of the noise.
Parameters¶
- x_newarray_like
2-D array with new training locations.
- y_newarray_like
1-D array with new training data observed at x_new. The length has to be the same as with x_new.
- fit(x, y, noise)[source]¶
Fit the Gaussian process to data y observed at training locations x with noise standard deviations noise.
Parameters¶
- xarray_like
2-D array with training locations.
- yarray_like
1-D array with training data observed at x. The length has to be the same as with x.
- noisefloat or array_like
Standard deviations of observational noise. If array_like, the length has to be the same as with y.
- property kernel¶
Kernel optimized using training data.
Returns¶
kernel : instance of a kernel class (depending on the underlying GPR implementation)
- property kernel_hyperparameters¶
Hyperparameters of kernel.
Returns¶
kernel_hyperparameters : ndarray of float
- property kernel_length_scales¶
Length scales of kernel.
Returns¶
kernel_length_scales : ndarray of float
- class ariane.lib.gaussian_process_regression.GaussianProcessRegressorImplementation[source]¶
Base class (interface) for implementations of Gaussian process regressors.
- class ariane.lib.gaussian_process_regression.implementations.sklearn.ScikitLearnGaussianProcessRegressor(kernel=None, *, noise=1e-10, optimizer='fmin_l_bfgs_b', n_restarts_optimizer=0, normalize_y=False, copy_X_train=True, random_state=None)[source]¶
Wrapper class for scikit-learn’s implementation of a Gaussian process regressor.
- add_data(x_new, y_new, noise=None)[source]¶
Add new data y_new measured at training locations x_new to the Gaussian process and update the posterior distribution. noise is again the standard deviation of the noise.
Parameters¶
- x_newarray_like
2-D array with new training locations.
- y_newarray_like
1-D array with new training data observed at x_new. The length has to be the same as with x_new.
- fit(x, y, noise=None)[source]¶
Fit the Gaussian process to data y observed at training locations x with noise standard deviations noise.
Parameters¶
- xarray_like
2-D array with training locations.
- yarray_like
1-D array with training data observed at x. The length has to be the same as with x.
- noisefloat or array_like
Standard deviations of observational noise. If array_like, the length has to be the same as with y.
- property kernel¶
Kernel optimized using training data.
Returns¶
kernel : instance of a kernel class (depending on the underlying GPR implementation)
- property kernel_hyperparameters¶
Hyperparameters of kernel.
Returns¶
kernel_hyperparameters : ndarray of float
- property kernel_length_scales¶
Length scales of kernel.
Returns¶
kernel_length_scales : ndarray of float
