A series of measurement experiments was conducted on the Rigetti M3 machine. Starting with qubit 6, qubit 7 was selected and the par formed a circuit. No gates other than those needed to compute the computation basis were used to execute and then measure the ciruit. Once this was completed the same was done for qubt 6 and 0, and for qubit 6 and qubit 1, and again unril qubit 40.
Here is the set of calibration matrices which result from the experiment.
To observe cross talk we look at when we prepare in $|10>$ and observe how much we measure in $|11>$. We also observer when we prepare in $|01>$ and measure in $|11>$.
We consider the difference in readout error.
For $p{i|j = 0} - p{i|j = 1}$
When we wish to obtain the probability that a measured state is prepared in a certain state $\psi$ which is measured in $\tilde{\psi}>$.
$P(|\psi> \rightarrow |\tilde{\psi}>)$
We look at prepared $|00>$ and observe the measured states $|00>$ and $|01>$ and compute the difference in probabilities.
$P(|00> \rightarrow |00>) - P(|00> \rightarrow |01>)$
$P(|00> \rightarrow |00>) - P(|00> \rightarrow |10>)$
To explore please a simple example https://ediparquantum.github.io/ReadoutNoiseCharacterisation/