optimize API reference#

Noise#

class optimize.noise.CorrelatedNoiseProcess#

Bases: optimize.noise.NoiseProcess

Trait.

class optimize.noise.GaussianProcess(kernel=None)#

Bases: optimize.noise.CorrelatedNoiseProcess

A noise kernel defined through a single GP and diagonal error terms with an additional “jitter” parameter. Each jitter parameter must be named “jitter_label” where label is the data label.

class optimize.noise.NoiseProcess#

Bases: object

A base noise process class defined through a covariance matrix. This class is not useful to instantiate on its own.

name#

The name of this noise process. Defaults to None.

Type

str, optional

class optimize.noise.UnCorrelatedNoiseProcess#

Bases: optimize.noise.NoiseProcess

Trait.

class optimize.noise.WhiteNoiseProcess#

Bases: optimize.noise.UnCorrelatedNoiseProcess

class optimize.kernels.CorrelatedNoiseKernel(par_names=None)#

Bases: optimize.kernels.NoiseKernel

Behaves as a trait for now.

class optimize.kernels.NoiseKernel(par_names=None)#

Bases: object

A base noise kernel class defined through a covariance matrix. This class is not useful to instantiate on its own.

par_names#

The parameter names for this kernel, optional.

Type

list of strings, optional

initialize(p0)#

Default wrapper to initialize the kernel, storing the parameters.

class optimize.kernels.QuasiPeriodic(par_names=None)#

Bases: optimize.kernels.StationaryNoiseKernel

A Quasiperiodic kernel. The hyperparameters may be called anything, but must be in the order of amplitude, exp length scale, period length scale, and period.

compute_cov_matrix(pars, x1, x2)#

Computes the QP kernel.

Parameters

pars (Parameters) – The parameters to use.

Returns

The covariance matrix K.

Return type

np.ndarray

class optimize.kernels.StationaryNoiseKernel(par_names=None)#

Bases: optimize.kernels.CorrelatedNoiseKernel

Noise kernel which only dependes on the relative difference between 2 values. Also has trait like behavior.

compute_dist_matrix(x1, x2)#

Computes the stationary distance matrix, D_ij = |x_i - x_j|.

Parameters

  • x1 (np.ndarray) – The first vector.

  • x2 (np.ndarray) – The second vector.

Returns

The stationary distance matrix D_ij

Return type

np.ndarray

class optimize.kernels.UnCorrelatedNoiseKernel(par_names=None)#

Bases: optimize.kernels.NoiseKernel

Behaves as a trait for now.