Adding optimizer: ExcitationSolve

[S-Erik]

Hey everyone,

I'd like to add the ExcitationSolve (code, preprint) optimizer to tequila. See a code example below using the implemented ExcitationSolve optimizer.

Open questions:

• Should I add a test in the tests folder?
• I am not sure if the return value of the ExciationSolve minimize function is acceptable (currently a SciPyResults object).
• I am not sure if I used the history member variable correctly.

Code example

import tequila as tq

tq.show_available_optimizers() # Lists now: excitationsolve      | excitationsolve

# define a molecule within an active space
active_orbitals = [1, 2, 5]
geometry = """
Li  0.0 0.0 0.0
H   0.0 0.0 1.6
"""
molecule = tq.quantumchemistry.Molecule(
    geometry=geometry, basis_set="6-31g", active_orbitals=active_orbitals, transformation="bravyi-kitaev"
)

# get the qubit hamiltonian
H = molecule.make_hamiltonian()

# create an k-UpCCGSD circuit of order k
U = molecule.make_ansatz(name="UpCCGSD")

# define the expectationvalue
E = tq.ExpectationValue(H=H, U=U)

# compute reference energies
fci = molecule.compute_energy("fci")

# optimize
result = tq.minimize(objective=E, method="excitationsolve", initial_values=0.0)

print("VQE : {:+2.8}f".format(result.energy))
print("FCI : {:+2.8}f".format(fci))

[kottmanj]

errors in test seem to be unrelated to tq, but an internal issue in excitation_solve

[S-Erik]

A few errors are similar to this:

/opt/hostedtoolcache/Python/3.9.24/x64/lib/python3.9/site-packages/excitationsolve/excitation_solve.py:7: in <module>
    parameter_variations: list | np.ndarray,
E   TypeError: unsupported operand type(s) for |: 'type' and 'type'

excitationsolve uses the | operator for type annotations. This operator is unknown to python3.9. python3.9 supports the Union type annotations.

Is it possible to use python3.10 or newer in the git runner or should I make excitationsolve compatible with python3.9?

[kottmanj]

Ah, thanks for digging that up.
It was on my todo list to update the test runners to 3.11 anyway.
Thanks!

[kottmanj]

Issue is on the openfermion end. Issues with numpy.int64 which is not recognized as int anymore. Will remove affected transformations for now.
(quantumlib/OpenFermion#1171)