We know that in Qiskit noise models modeling specific quantum machine prototypes exist and can be inserted into the code. However, are these models fixed in nature or they continuously improve themselves using machine learning techniques so that they can provide results ideally depicting a prototype.
Also which is a better way of getting correct results. Designing a circuit in circuit composer and executing on a selected IBM machine or writing a code for the circuit in Qiskit and executing the circuit on QASM simulator with a noise model for a IBM quantum machine inserted into the code.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
I would think the noise model are based off the calibration data of the real hardware. I am not sure how often IBM updated them. Hopefully someone from the IBM team can provide more details.
In term your second question/point. After designing my quantum algorithm/circuit, I would first test it on the simulator without any noise and see how it works first. Making sure that everything works the way it supposed too, at least on a small scale. Then after seeing that it works on simulator on a small scale, I can now start adding noise to the simulator to see how it performs. After all this, I would then perform circuit run on the real hardware and perhaps doing it with error mitigation techniques.
-
$\begingroup$ but dont you think a quantum computer noise model execution on simulator should give a more accurate result as its designed as per specific quantum machine, while the quantum computer execution may show more deviation in results due to dynamic changes the hardware may be experiencing due to environmental and other factors affecting its accuracy. $\endgroup$– parthDec 7, 2020 at 8:15