groentelepel / qiskit-quantum-knn Goto Github PK

When computing fidelities/contrasts, there is a step where we map the states of the computational basis (e.g. '10010101') back to integers using

index_state = int(comp_state, 2)

However, I read in the documentation of Qiskit that the ordering they use for tensor product states is the opposite of most physics textbooks:

When representing the state of a multi-qubit system, the tensor order used in Qiskit is different than that used in most physics textbooks. Suppose there are 𝑛 qubits, and qubit 𝑗 is labeled as 𝑄𝑗. Qiskit uses an ordering in which the 𝑛th qubit is on the left side of the tensor product, so that the basis vectors are labeled as 𝑄𝑛−1⊗⋯⊗𝑄1⊗𝑄0.

For example, if qubit zero is in state 0, qubit 1 is in state 0, and qubit 2 is in state 1, Qiskit would represent this state as |100⟩, whereas many physics textbooks would represent it as |001⟩.

I think this means we might have to change the code to

index_state = int(comp_state[::-1], 2)

I played around with this a little bit in a juptyer notebook, where I tried to explicitly compute the fidelities of some sample states, and found that without this change, a lot of them get swapped around. Reversing the order as above seems to fix the issue.

If what I'm saying is correct, we'd have to make 2 changes:

Running the QKneighborsClassifier using

n_variables = 4
n_train_points = 4
n_test_points = 1

Results in an error:

File "qiskit_quantum_knn/qknn/qkneighborsclassifier.py", line 266, in get_all_contrasts
    num_qubits = len(list(all_counts[0].keys())[0]) - 2
KeyError: 0

Since the qiskit aqua is depricated now, this code is not working anymore. Can somebody please help me in implementing this QKNN with latest qiskit libraries.

Hello !
It would be great if you released a new patched version so that your library can be easily accessed.
Now it doesn't work because of aqua.
Have a nice day!