qml.evolve: dispatch function

[AlbertMitjans]

This PR does the following:

1. Use singledispatch to dispatch qml.evolve into parametrized_evolution and evolution, each with its necessary arguments.
2. Add a -1 when using qml.evolve with an Operator, to be consistent with ParametrizedEvolution.
3. Add a deprecation warning in Evolution.__init__ redirecting users to qml.evolve.
4. Skip the warning in Evolution.__init__ when using qml.evolve.
5. Remove unused ParametrizedEvolution.time argument.

[codecov]

Codecov Report

Merging #3706 (eca547f) into master (876bac8) will decrease coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #3706      +/-   ##
==========================================
- Coverage   99.74%   99.73%   -0.01%     
==========================================
  Files         328      328              
  Lines       28662    28695      +33     
==========================================
+ Hits        28588    28619      +31     
- Misses         74       76       +2     

Impacted Files	Coverage Δ
pennylane/devices/default_qubit.py	99.69% <ø> (-0.01%)	⬇️
pennylane/pulse/parametrized_evolution.py	95.45% <ø> (-0.11%)	⬇️
pennylane/ops/functions/evolve.py	100.00% <100.00%> (ø)
pennylane/ops/op_math/evolution.py	100.00% <100.00%> (ø)
pennylane/ops/qubit/state_preparation.py	100.00% <100.00%> (ø)
pennylane/devices/default_qubit_tf.py	98.03% <0.00%> (-1.97%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[lillian542]

I'm not sure about adding t: for the Evolution operator, it seems to suggest that the coefficient in exp(ixG) is time, as it is set by t=3 in the function call. The Evolution class is described in the docstring as implementing exp(ixG) rather than exp(iHt) because it should be a more general thing that also encompasses, for example, general rotation operators:

>>> theta = np.array(1.23)
>>> op = Evolution(qml.PauliX(0), -0.5 * theta)
>>> qml.math.allclose(op.matrix(), qml.RX(theta, wires=0).matrix())
True

So I'm not convinced it makes things clearer, though I do like having the option to set the param in qml.evolve.

[AlbertMitjans]

I'm not sure about adding t: for the Evolution operator, it seems to suggest that the coefficient in exp(ixG) is time

A rotation operator is considered an evolution? When googling "evolution operator" all the links talk about time evolution.

[lillian542]

A rotation operator is considered an evolution? When googling "evolution operator" all the links talk about time evolution.

This was also my initial response, and I think the name is not ideal. The idea with the operator was a generalized exp(ixG), without regard for whether something was being evolved 'in time' or in some other parameter.

On the original shortcut ticket it was called Rotation, but that sounds like it's not for time evolution. So it ended up being called Evolution, which sort of sounds like it's not for rotation (or anything else but time evolution). The description and examples in the docstring attempt to clarify that it is meant as a more general mathematical object exp(ixG). But we couldn't come up with a good general name for exp(ixG). I feel like this could be improved, but I also don't have any good suggestions.

[lillian542]

Looks good to me! 🎉