You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently the QX emulator supports the depolarizing error model (see Mike and Ike, chapter 8.3.4, formula 8.100). It is actually a so-called symmetric depolarizing error model with p=1-px-py-pz.
It is relatively straightforward to add a bit-flip error model and a phase-flip error model (see Mike and Ike, chapter 8.3.3, formula 8.94), which inserts not X,Y and Z errors, but only X or only Z errors with probability px or pz, or asymmetric error models with different values for px, py and pz, see paper
An asymmetric model is probably the easiest to add. This will require three parameters px, py and pz. By setting some parameters to zero, the user can then also enforce a bit-flip or phase-flip error channel. My suggestion would be (if that is possible) that the current depolarizing model will take three parameters as an input (px, py and pz). When there is just one parameter, it is assumed that a symmetric model is uses, otherwise three parameters are required.
This is also a first step towards more realistic error models, like amplitude and phase damping, which can be approximated by Pauli twirling (see same paper as above). This requires two more parameters (T1 and T2* times and duration of the gates, information that could for instance be provided by annotations to QX.
The text was updated successfully, but these errors were encountered:
Currently the QX emulator supports the depolarizing error model (see Mike and Ike, chapter 8.3.4, formula 8.100). It is actually a so-called symmetric depolarizing error model with p=1-px-py-pz.
It is relatively straightforward to add a bit-flip error model and a phase-flip error model (see Mike and Ike, chapter 8.3.3, formula 8.94), which inserts not X,Y and Z errors, but only X or only Z errors with probability px or pz, or asymmetric error models with different values for px, py and pz, see paper
An asymmetric model is probably the easiest to add. This will require three parameters px, py and pz. By setting some parameters to zero, the user can then also enforce a bit-flip or phase-flip error channel. My suggestion would be (if that is possible) that the current depolarizing model will take three parameters as an input (px, py and pz). When there is just one parameter, it is assumed that a symmetric model is uses, otherwise three parameters are required.
This is also a first step towards more realistic error models, like amplitude and phase damping, which can be approximated by Pauli twirling (see same paper as above). This requires two more parameters (T1 and T2* times and duration of the gates, information that could for instance be provided by annotations to QX.
The text was updated successfully, but these errors were encountered: