searchep.gpr#

Perform gaussian processes with GPflow

searchep.gpr.gp_2d_diff_kappa(training_data: Data)#

2D gaussian process model for eigenvalue difference with respect to kappa

GPflow is used to make a prediction model for the eigenvalue (ev) difference with respect to kappa. Due to complex kappa- and eigenvalues it is a 2D on 2D model.

Parameters:

  • training_data (data.Data) – Class which contains all scale-, kappa- and eigenvalues

  • ev (#) –

  • EP (# 2D complex array which contains the two eigenvalues belonging to the) –

  • kappa (#) –

  • values (# 1D array which contains the complex kappa) –

Returns:

Returns a 2D GPR model for the complex eigenvalue difference squared with respect to kappa and a 1D array which contains the kernel eigenvalues of the input space.

Return type:

(gpflow.models.GPR, np.ndarray)

searchep.gpr.gp_2d_sum_kappa(training_data: Data)#

2D gaussian process model for eigenvalue sum with respect to kappa

GPflow is used to make a prediction model for the eigenvalue (ev) sum with respect to kappa. Due to complex kappa- and eigenvalues it is a 2D on 2D model.

Parameters:

  • training_data (data.Data) – Class which contains all scale-, kappa- and eigenvalues

  • ev (#) –

  • EP (# 2D complex array which contains the two eigenvalues belonging to the) –

  • kappa (#) –

  • values (# 1D array which contains the complex kappa) –

Returns:

Returns a 2D GPR model for the complex eigenvalue sum with respect to kappa and a 1D array which contains the kernel eigenvalues of the input space.

Return type:

(gpflow.models.GPR, np.ndarray)

searchep.gpr.gp_create_matern52_model(kappa: ndarray, validation_data: ndarray, training_data: Data)#

2D gaussian process model with matern52 kernel

GPflow is used to make a 2D prediction model with the matern52 kernel.

Parameters:

  • kappa (np.ndarray) – 2D array which contains the real and the imaginary part of all kappa values

  • validation_data (np.ndarray) – 2D array which contains usually the real and imaginary part of the validation data

  • training_data (data.Data) – Class which contains all scale-, kappa- and eigenvalues

Returns:

Returns a 2D GPR model created with the matern52 kernel and a 1D array which contains the kernel eigenvalues of the input space.

Return type:

(gpflow.models.GPR, np.ndarray)

searchep.gpr.gp_create_rbf_model(kappa: ndarray, validation_data: ndarray, training_data: Data)#

2D gaussian process model with rbf kernel

GPflow is used to make a 2D prediction model with the rbf kernel.

Parameters:

  • kappa (np.ndarray) – 2D array which contains the real and the imaginary part of all kappa values

  • validation_data (np.ndarray) – 2D array which contains usually the real and imaginary part of the validation data

  • training_data (data.Data) – Class which contains all scale-, kappa- and eigenvalues

Returns:

Returns a 2D GPR model created with the rbf kernel and a 1D array which contains the kernel eigenvalues of the input space.

Return type:

(gpflow.models.GPR, np.ndarray)

searchep.gpr.gp_diff_angle(ev, phi)#

Gaussian process model for eigenvalue difference with respect to the angle

GPflow is used to make a prediction model for the eigenvalue (ev) difference with respect to the angle (phi). Due to complex eigenvalues there are two models. One for the real and one for the imaginary part. Only works for 2D matrix.

Parameters:

  • ev (np.ndarray) – 2D complex array which contains the eigenvalues of the 2D matrix

  • phi (np.ndarray) – 1D array which contains the related angles

Returns:

Returns two GPR models. One GPR model for the real part of the eigenvalue difference and one GPR model for the imaginary part of the eigenvalue difference

Return type:

(gpflow.models.GPR, gpflow.models.GPR)