How do I bound parameters in RawFeatureVector with ParameterizedInitialize in Qiskit?

Question

I'm not sure how to bind the parameters in the Qiskit RawFeatureVector circuit. Here is my code:

feature_map = RawFeatureVector(feature_dimension = num_features)
feature_map = feature_map.assign_parameters(np.array([1,1]) / np.sqrt(2))
sampler = Sampler()
fidelity = ComputeUncompute(sampler=sampler)
kernel = FidelityQuantumKernel(fidelity=fidelity, feature_map=feature_map)
svm = OneClassSVM(kernel = kernel.evaluate, verbose=True, nu=outliers_fraction)
svm.fit(X)

I am probably missing something very obvious, but I think I bound the parameters when I assigned the paramters. Now when I run this with some data, I get the following error traceback.

QiskitError                               Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_8160/3101503440.py in <module>
      4     dataset_count = dataset_count + 1
      5     print("For dataset: {}".format(dataset_count))
----> 6     Algorithm2(X, y, 1, outliers_fraction=outliers_fraction)
      7     break

~\AppData\Local\Temp/ipykernel_8160/1216541068.py in Algorithm2(X, y, shots, outliers_fraction, num_features, seed, predict, supervised)
     17     if supervised:
     18         X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state=seed)
---> 19         svm.fit(X_train, y_train)
     20         y_pred = svm.predict(X_test)
     21         # TODO save to Matrix

~\anaconda3\lib\site-packages\sklearn\svm\_classes.py in fit(self, X, y, sample_weight, **params)
   1374 
   1375         """
-> 1376         super().fit(X, np.ones(_num_samples(X)),
   1377                     sample_weight=sample_weight, **params)
   1378         self.offset_ = -self._intercept_

~\anaconda3\lib\site-packages\sklearn\svm\_base.py in fit(self, X, y, sample_weight)
    224 
    225         seed = rnd.randint(np.iinfo('i').max)
--> 226         fit(X, y, sample_weight, solver_type, kernel, random_seed=seed)
    227         # see comment on the other call to np.iinfo in this file
    228 

~\anaconda3\lib\site-packages\sklearn\svm\_base.py in _dense_fit(self, X, y, sample_weight, solver_type, kernel, random_seed)
    264             # TODO: add keyword copy to copy on demand
    265             self.__Xfit = X
--> 266             X = self._compute_kernel(X)
    267 
    268             if X.shape[0] != X.shape[1]:

~\anaconda3\lib\site-packages\sklearn\svm\_base.py in _compute_kernel(self, X)
    394             # in the case of precomputed kernel given as a function, we
    395             # have to compute explicitly the kernel matrix
--> 396             kernel = self.kernel(X, self.__Xfit)
    397             if sp.issparse(kernel):
    398                 kernel = kernel.toarray()

~\anaconda3\lib\site-packages\qiskit_machine_learning\kernels\fidelity_quantum_kernel.py in evaluate(self, x_vec, y_vec)
    119         if is_symmetric:
    120             left_parameters, right_parameters, indices = self._get_symmetric_parameterization(x_vec)
--> 121             kernel_matrix = self._get_symmetric_kernel_matrix(
    122                 kernel_shape, left_parameters, right_parameters, indices
    123             )

~\anaconda3\lib\site-packages\qiskit_machine_learning\kernels\fidelity_quantum_kernel.py in _get_symmetric_kernel_matrix(self, kernel_shape, left_parameters, right_parameters, indices)
    210         Given a set of parameterization, this computes the kernel matrix.
    211         """
--> 212         kernel_entries = self._get_kernel_entries(left_parameters, right_parameters)
    213         kernel_matrix = np.ones(kernel_shape)
    214 

~\anaconda3\lib\site-packages\qiskit_machine_learning\kernels\fidelity_quantum_kernel.py in _get_kernel_entries(self, left_parameters, right_parameters)
    232                 right_parameters,
    233             )
--> 234             kernel_entries = np.real(job.result().fidelities)
    235         else:
    236             # trivial case, only identical samples

~\anaconda3\lib\site-packages\qiskit\primitives\primitive_job.py in result(self)
     48         """Return the results of the job."""
     49         self._check_submitted()
---> 50         return self._future.result()
     51 
     52     def cancel(self):

~\anaconda3\lib\concurrent\futures\_base.py in result(self, timeout)
    436                     raise CancelledError()
    437                 elif self._state == FINISHED:
--> 438                     return self.__get_result()
    439 
    440                 self._condition.wait(timeout)

~\anaconda3\lib\concurrent\futures\_base.py in __get_result(self)
    388         if self._exception:
    389             try:
--> 390                 raise self._exception
    391             finally:
    392                 # Break a reference cycle with the exception in self._exception

~\anaconda3\lib\concurrent\futures\thread.py in run(self)
     50 
     51         try:
---> 52             result = self.fn(*self.args, **self.kwargs)
     53         except BaseException as exc:
     54             self.future.set_exception(exc)

~\anaconda3\lib\site-packages\qiskit\algorithms\state_fidelities\compute_uncompute.py in _run(self, circuits_1, circuits_2, values_1, values_2, **options)
    126         """
    127 
--> 128         circuits = self._construct_circuits(circuits_1, circuits_2)
    129         if len(circuits) == 0:
    130             raise ValueError(

~\anaconda3\lib\site-packages\qiskit\algorithms\state_fidelities\base_state_fidelity.py in _construct_circuits(self, circuits_1, circuits_2)
    186                 parametrized_circuit_2 = circuit_2.assign_parameters(parameters_2)
    187 
--> 188                 circuit = self.create_fidelity_circuit(
    189                     parametrized_circuit_1, parametrized_circuit_2
    190                 )

~\anaconda3\lib\site-packages\qiskit\algorithms\state_fidelities\compute_uncompute.py in create_fidelity_circuit(self, circuit_1, circuit_2)
     91             circuit_2.remove_final_measurements()
     92 
---> 93         circuit = circuit_1.compose(circuit_2.inverse())
     94         circuit.measure_all()
     95         return circuit

~\anaconda3\lib\site-packages\qiskit\circuit\library\blueprintcircuit.py in inverse(self)
    132         if not self._is_built:
    133             self._build()
--> 134         return super().inverse()
    135 
    136     def __len__(self):

~\anaconda3\lib\site-packages\qiskit\circuit\quantumcircuit.py in inverse(self)
    605 
    606         for instruction in reversed(self._data):
--> 607             inverse_circ._append(instruction.replace(operation=instruction.operation.inverse()))
    608         return inverse_circ
    609 

~\anaconda3\lib\site-packages\qiskit\circuit\instruction.py in inverse(self)
    364                 and an inverse has not been implemented for it.
    365         """
--> 366         if self.definition is None:
    367             raise CircuitError("inverse() not implemented for %s." % self.name)
    368 

~\anaconda3\lib\site-packages\qiskit\circuit\instruction.py in definition(self)
    235         """Return definition in terms of other basic gates."""
    236         if self._definition is None:
--> 237             self._define()
    238         return self._definition
    239 

~\anaconda3\lib\site-packages\qiskit_machine_learning\circuit\library\raw_feature_vector.py in _define(self)
    168                 cleaned_params.append(complex(param))
    169             else:
--> 170                 raise QiskitError("Cannot define a ParameterizedInitialize with unbound parameters")
    171 
    172         # normalize

QiskitError: 'Cannot define a ParameterizedInitialize with unbound parameters'

Thanks in advance!

Answer 1

Well, I'm facing a similar problem while using the RealFeatureVector with NN and some optimizer. This is my interpretation: In both cases the circuit needs to be transpiled/decomposed to be evaluated in each context (in my case was to build a gradient of the circuit for the sake of the optimizer). However, the RealFeatureVector is different from any other feature-map; that is, the circuit can't be constructed till the parameters are bound. The gates are added in a specific way to achieve the required initialization value.

Accordingly, I would recommend to try the following: 1- Change assign_parameters to bind_parameters 2- I can't see you using any parameters, so you may use a normal QuantumCircuit and then initialize it by the desired values.

regards, Ahmed