Skip to content

Filters

mllabs.DataFilter

Base class for data filters used with :class:~mllabs.collector.SHAPCollector.

Subclasses implement :meth:_select to return a row index array. The filter is applied to all arrays in the input data_dict identically.

__call__(data_dict)

Apply the filter to every array in data_dict.

Parameters:

Name Type Description Default
data_dict dict

{key: DataWrapper} mapping.

required

Returns:

Name Type Description
dict

Filtered {key: DataWrapper} with rows selected by

meth:_select.

mllabs.RandomFilter

Bases: DataFilter

Randomly subsample rows from a data dict.

Parameters:

Name Type Description Default
n int

Absolute number of rows to sample. Mutually exclusive with frac.

None
frac float

Fraction of rows to sample (0–1). Mutually exclusive with n.

None
random_state int

Random seed for reproducibility.

None

mllabs.IndexFilter

Bases: DataFilter

Select rows whose index values appear in a provided index array.

Parameters:

Name Type Description Default
index array - like

The set of index values to keep.

required