The construct_gaussian_process function#

Aliases#

halerium.core.template.construct_gaussian_process
construct_gaussian_process(name, shape, kernel, axis=-1, mean=None, variable_class=<class 'halerium.core.variable.static_variable.StaticVariable'>)#

Construct a Gaussian process based on a shape and a kernel. The Gaussian process will be constructed by convolving a standard normal variable with the procided kernel along the specified axis or axes. The Gaussian process will be constructed as a Variable or StaticVariable depending on the provided variable_class. This variable will contain the standard normal variable as well as the kernel. This special construction will be treated differently when posterior graphs are built to keep the convolution properties.

Parameters:

  • name (str) – The name of the constructed variable.

  • shape (tuple) – The target shape of the Gaussian process

  • kernel (Operator, numpy.ndarray or tuple or list) – The kernel. Must not depend on Variable instances, only on StaticVariable instances. If a tuple or list is provided it must contain operators and the axis argument must be a list or tuple of the same length.

  • axis (int, tuple or list, optional) – The axis or axes on which are to be convolved with the kernel. If a list or tuple is provided as axis and kernel is a list or tuple as well the kernels are applied to their corresponding axis. If a list or tuple is provided as axis and kernel is a single operator the same kernel is applied to all axis entries. The default is -1.

  • mean (Operator , optional) – an additional mean to provide to the Gaussian process. The default is None.

  • variable_class (type, optional) – The desired class for the Gaussian process. Either StaticVariable or Variable. The default is StaticVariable.

Returns:

The constructed Gaussian process of the specified variable_class type.

Return type:

Variabe or StaticVariable