How to get angles optimized by classical optimizer from QisKit's QAOA Module?

Question

I have the following simple optimization in QAOA:

from qiskit_optimization.algorithms import MinimumEigenOptimizer

# from qiskit_aer import Aer
from qiskit.algorithms.minimum_eigensolvers import QAOA
from qiskit.algorithms.optimizers import COBYLA
from qiskit.primitives import Sampler

n_qubits = len(G.nodes())
problem = QuadraticProgram()
_ = [problem.binary_var("x{}".format(i)) for i in range(n_qubits)]
problem.maximize(
    linear=nx.adjacency_matrix(G).dot(np.ones(n_qubits)),
    quadratic=-nx.adjacency_matrix(G),
)

meo = MinimumEigenOptimizer(QAOA(sampler=Sampler(), optimizer=COBYLA(maxiter=100)))
result = meo.solve(problem)
print(result.prettyprint())
print("\ndisplay the best 5 solution samples")
for sample in result.samples[:5]:
    print(sample)

I want to get the actual angles found by QAOA, the $\beta, \gamma$. How do I get these from the results of this algorithm?

I'm not seeing it in the docs: https://qiskit.org/documentation/stubs/qiskit.algorithms.minimum_eigensolvers.QAOA.html#qiskit.algorithms.minimum_eigensolvers.QAOA

Answer 1

To get optimized parameters from QAOA you can also do this. The MinimumEigenOptimizer returns a MinimumEigenOptimizationResult which has a field min_eigen_solver_result which is as the API ref linked states the result obtained from the underlying algorithm. Now QAOA extends SamplingVQE and provides an identical result object, a SamplingVQEResult. The final β,γ can be in that field optimal_point which is just the list of floats the optimizer was working with, or in optimal_parameter which is a dictionary of the β,γ parameters to the values.

Your code sample did not run, I edited it to add a G taken from the Optimization MaxCut tutorial and other imports.

from qiskit_optimization import QuadraticProgram
from qiskit_optimization.algorithms import MinimumEigenOptimizer

from qiskit.algorithms.minimum_eigensolvers import QAOA
from qiskit.algorithms.optimizers import COBYLA
from qiskit.primitives import Sampler
import numpy as np

import networkx as nx
n = 4
G = nx.Graph()
G.add_nodes_from(np.arange(0, n, 1))
elist = [(0, 1, 1.0), (0, 2, 1.0), (0, 3, 1.0), (1, 2, 1.0), (2, 3, 1.0)]
G.add_weighted_edges_from(elist)

n_qubits = len(G.nodes())
problem = QuadraticProgram()
_ = [problem.binary_var("x{}".format(i)) for i in range(n_qubits)]
problem.maximize(
    linear=nx.adjacency_matrix(G).dot(np.ones(n_qubits)),
    quadratic=-nx.adjacency_matrix(G),
)

meo = MinimumEigenOptimizer(QAOA(sampler=Sampler(), optimizer=COBYLA(maxiter=100)))
result = meo.solve(problem)
print(result.prettyprint())
print("\ndisplay the best 5 solution samples")
for sample in result.samples[:5]:
    print(sample)

# Print the final QAOA parameters

print(result.min_eigen_solver_result.optimal_point)
print(result.min_eigen_solver_result.optimal_parameters)

the extra prints I added printed this for me

[5.60276426 4.22978775]
{ParameterVectorElement(β[0]): 5.602764261565667, ParameterVectorElement(γ[0]): 4.229787753395598}

Answer 2

You can pass a callback function to QAOA. This callback can access the intermediate data at each optimization step. This data includes the optimizer parameters for the ansatz:

def callback(evaluation_count, optimizer_parameters, estimated_value, metadata):
    print(optimizer_parameters)

meo = MinimumEigenOptimizer(QAOA(sampler=Sampler(), optimizer=COBYLA(maxiter=100), callback=callback))