0
$\begingroup$

I've been running VQE experiments locally using both the statevector_simulator and qasm_simulator, however I've noticed that the runtimes for the qasm_simulator are significantly longer than that of statevector_simulator. According to this answer the statevector_simulator produces ideal/deterministic results, whereas qasm_simulator produces non-ideal/probabilistic results. However I don't see why the simulation times should differ significantly($>\times 100$) if the number of shots=1 for the qasm_simulator even if a NoiseModel or measurement_error_mitigation_cls=CompleteMeasFitter is used.

Is this in fact expected and how is it explained?

$\endgroup$
0
$\begingroup$

Note that if you use qasm_simulator with shots=1, this doesn't tell you anything! Unless IBM automatically switch your simulator to statevector_simulator if you specify the shots =1 when calling qasm_simulator.

Now in term of add noise model and doing measurement_error_mitigation. To be able to do this, you need to generate $2^n$ basis states and measure each of them to create a calibration matrix. It is an expensive procedure since you have many more quantum circuits to execute. Each circuit requires a certain amount of shots, up to 8192 shots. In statevector_simulator, you don't need to do this as there is no noise.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.