pyemma.coordinates.data.CustomFeature

class pyemma.coordinates.data.CustomFeature(func=None, *args, **kwargs)

A CustomFeature is the base class for user-defined features. If you want to implement a new fancy feature, derive from this class, calculate the quantity of interest in the map method and return it as an ndarray.

If you have defined a map function that should be classed, you don’t need to derive a class, but you can simply pass a function to the constructor of this class

Parameters:

• func (function) – will be invoked with given args and kwargs on mapping traj
• args (list of positional args (optional) passed to func) –
• kwargs (named arguments (optional) passed to func) –

Examples

We define a feature that transforms all coordinates by \(1 / x^2\):

>>> from pyemma.coordinates import source
# define a function which transforms the coordinates of the trajectory object
>>> my_feature = CustomFeature(lambda x: 1.0 / x.xyz**2)
>>> reader = load('traj.xtc', top='my_topology.pdb')
# pass the feature to the featurizer and transform the data
>>> reader.featurizer.add_custom_feature(my_feature)
>>> data = reader.get_output()

__init__(func=None, *args, **kwargs)

Methods

__init__([func])
describe()
map(traj)