Many gates are not available on a real computer and therefore the circuit must be transpiled into a specific set of gates. I have seen this equation below which is used to to determine the 'cost' of a quantum circuit:

Cost = 10NCNOT + Nother

NCNOT : number of CNOT gates
Nother : number of other gates

I am wondering which other considerations are not captured by this equation? Especially given that we live in the Noisy Intermediate-Scale Quantum (NISQ) technology era.

It has obviously left out the number of qubits in a circuit which contributes massively to noise. What else should it take into account though? very curious

  • 1
    $\begingroup$ Are you sure the cost isn't defined as $10 \cdot N_{CNOT} + N_{other}$? $\endgroup$
    – epelaez
    Jun 28, 2021 at 16:54
  • $\begingroup$ Yes, you are right sorry. Typo $\endgroup$
    – John
    Jun 28, 2021 at 16:57
  • $\begingroup$ @epelaaez Do you have any useful sources on the cost equation? I have been trying to find but cannot seem to. $\endgroup$
    – John
    Jun 28, 2021 at 16:59
  • $\begingroup$ I've seen that's the cost function that IBM uses in their challenges, but I haven't seen it stated in some more "formal" place $\endgroup$
    – epelaez
    Jun 28, 2021 at 17:04
  • 1
    $\begingroup$ I'd be wary of linear formulas like this. Let's say that all 2 qubit gates have $10\times$ as many errors as single-qubit gates, where we call the single-qubit error gates $\eta$. The total success probability will then be $(1-\eta)^{N_{\mathrm{other}}}(1-10\eta)^{N_{\mathrm{CNOT}}}\approx 1-\eta(N_o+10N_C)+\cdots$. Obviously to first order the error looks like it scales with $N_o+10N_C$, but when you add enough gates to have $\eta(N_o+10N_C)\sim 1$ then all of the other terms become necessary and the relationship breaks down. $\endgroup$ Jun 29, 2021 at 13:01


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.