I'm trying to implement my first variational regressor using qiskit. I would like to understand how the interpret of a CircuitQNN works. I need to define an interpret that works on the amplitudes of a certain set of measured basis states. Considering this I'm trying to get the counts obtained by executing the circuit for S shots and then perform some operations using these values to finally obtain the output of my regressor.
I saw the Qiskit's tutorial example (https://qiskit.org/documentation/machine-learning/tutorials/02_neural_network_classifier_and_regressor.html#Regression) in which they use the parity to get the class of an instance. But, I think they compute the parity for the bitstring measured at each shot, here I would like to execute all the S shots and then on the counts perform some operations.