The DataLinker class#

Aliases#

halerium.core.DataLinker

halerium.core.data_linker.DataLinker

class DataLinker(n_data=1)#

The DataLinker holds data in the form of numpy arrays which are to be linked to specific Variables in a Graph.

Additionally, the DataLinker dictates the size of the data-dimension. All non-static variables will be scaled up according to n_data. For example, a Variable with a shape of (3, 4) will be represented in the model by an array of shape (n_data, 3, 4).

DataLinker instances are therefore a key ingredient when creating a model.

Holds links of data to variables in a graph.

Parameters:: n_data (int) – The number of examples in data. The default is 1.

add_links_from(data)#

Add links from data.

Parameters:: data (dict) – A dictionary with Variable instances as keys and array-likes as values.

assert_has_compatible_shape(variable, array, name_of_array)#

Assert variable and array have compatible shape.

Parameters:

variable – The variable with given shape.
array – The array to check shape.
name_of_array (str) – The name of the array in the potential error message.

Raises:

ValueError : – If shape of variable and array are incompatible.

copy(deep=False)#

Make a copy.

Returns a copy of the DataLinker instance which holds the same data links.

Note that unless deep=True, the `numpy.ndarray`s holding the data are not copied but referenced. So in-place manipulations of the arrays will change both the original and the copied data_linker.

Parameters:: deep (bool, optional) –
Returns:: copy_instance – The copy of the data linker.
Return type:: DataLinker

get_data_and_mask(variable_or_global_name)#

Get data and mask.

Get the data and mask, if present, for a given variable, else return None.

Parameters:

variable_or_global_name (VariableBase or string) – either the variable instance or its global name

Returns:

data (np.ndarray, None) – Ff the DataLinker contains a link for the given variable, the data are returned. Otherwise, the return is None.
mask (np.ndarray, None) – If the DataLinker contains a link for the given variable and the data have masked entries, the boolean mask is returned. Otherwise, the return is None.

get_full_shape(variable)#

Get full shape of variable.

Parameters:: variable (VariableBase) – The variable to compute the full shape for.
Returns:: shape – The full shape of the variable.
Return type:: tuple

has_data(variable_or_global_name)#

Inquire whether a given variable has data.

Parameters:: variable_or_global_name (str, VariableBase) – Either the variable instance or its global name.
Returns:: Whether the variable has a data.
Return type:: bool

is_fully_determined_by_data(variable_or_global_name)#

Inquire whether a given variable is fully determined by data linked to the DataLinker. A variable is fully determined, if the data linker contains a link without any masked entries in the data.

Parameters:: variable_or_global_name (VariableBase or string) – either the variable instance or its global name
Returns:: bool
Return type:: True if fully determined, otherwise False

items()#

Returns:: The items of the link registry.
Return type:: items

keys()#

Returns:: The keys of the link registry (global names of variables).
Return type:: keys

link(variable, data, mask=None)#

Link data to variable.

The method to link data to a halerium variable. Additionally, to the data the user can provide a mask, that flags unknown values in the data array.

Parameters:

variable (VariableBase) – The Variable instance to which the data belong.
data (array-like) – The data to be linked to the variable. It will be cast to a numpy.ndarray of dtype float. Non-finite values will be masked automatically, but a mask can also be provided explicitly.
mask (numpy.ndarray, None) – The mask to apply to data. Either None, or a numpy.ndarray of dtype bool. If provided, the indices containing True will be masked in the data array.

property n_data#

property registry#

A dictionary containing the linked data assigned to variable names.

Warning: Although a new dict is created and returned, the returned dict values are numpy arrays which are mutable. Thus changing values in these arrays in this instance’s _link_registry will change these values in the returned dict as well (and vice versa).

Returns:: A shallow copy of the instance’s link registry.
Return type:: registry

reset()#: Reset the DataLinker instance. All links are deleted.

to_compatible_shape(variable, array)#

Cast array to shape compatible with variable.

Parameters:

variable (VariableBase) – The variable to which the array shape to adjust.
array (numpy.ndarray) – The array whose shape is to be adjusted.

Returns:

array – The array with possibly adjusted shape compatible with variable.

Return type:

np.ndarray

unlink(variable_or_global_name)#

Remove the data link for a specific variable.

Parameters:: variable_or_global_name (VariableBase, str) – Either the variable instance or its global name.

update(other)#

Update self with other.

Update self with data from other overwriting existing keys.

Parameters:: other (dict, DataLinker,) – The other data to update self with.

values()#

Returns:: The values of the link registry (pairs of variable data and mask).
Return type:: values