DiscretizedSpace.element

DiscretizedSpace.element(inp=None, copy=None, **kwargs)

Create an element from inp or from scratch.

Parameters

inpoptional

Input used to initialize the new element. The following options are available:

• None: an empty element is created with no guarantee of its state (memory allocation only).

• array-like: an element wrapping a tensor is created, where a copy is avoided whenever possible. This usually requires correct shape, dtype and impl if applicable. See the element method of tspace for more information.

  If any of these conditions is not met, a copy is made.

• callable: a new element is created by sampling the function using point_collocation.

kwargs :

Additional arguments passed on to point_collocation when called on inp, in the form point_collocation(inp, points, **kwargs). This can be used e.g. for functions with parameters.

Returns

elementDiscretizedSpaceElement

The discretized element, calculated as point_collocation(inp) or tspace.element(inp), tried in this order.

Examples

Elements can be created from array-like objects that represent an already discretized function:

>>> space = odl.uniform_discr(-1, 1, 4)
>>> space.element([1, 2, 3, 4])
uniform_discr(-1.0, 1.0, 4).element([ 1.,  2.,  3.,  4.])
>>> vector = odl.rn(4).element([0, 1, 2, 3])
>>> space.element(vector)
uniform_discr(-1.0, 1.0, 4).element([ 0.,  1.,  2.,  3.])

On the other hand, non-discretized objects like Python functions can be discretized “on the fly”:

>>> space.element(lambda x: x * 2)
uniform_discr(-1.0, 1.0, 4).element([-1.5, -0.5, 0.5, 1.5])

This works also with parameterized functions, however only through keyword arguments (not positional arguments with defaults):

>>> def f(x, c=0.0):
...     return np.maximum(x, c)
...
>>> space = odl.uniform_discr(-1, 1, 4)
>>> space.element(f, c=0.5)
uniform_discr(-1.0, 1.0, 4).element([ 0.5 ,  0.5 ,  0.5 ,  0.75])