Skip to content

Measures of noise strength

Pavithran S Iyer edited this page Apr 16, 2018 · 15 revisions

Several measures for noise strength have been defined in chflow. The below table summarizes the names and descriptions of the noise metrics.

name Description
dnorm Diamond distance of the input channel from the identity channel
infid 1 - (Fidelity of the input channel's Choi matrix to the Bell state)
frb L2 - distance of the input channel's Choi matrix to the Bell state
trn Trace norm distance between the input channel's Choi matrix and the Bell state
nonu 1 - Unitarity of the input channel, as defined in
bd Bures distance between the input channel and the identity channel
errp A measure of error probability of the input channel, proposed in
entropy Von Neumann entropy of the input channel's Choi matrix
uhl Uhl-Josza fidelity of the input channel to the identity channel
np1 L2 norm of the difference between the channel's Chi matrix and it's twirled approximation.
np2 Least fidelity between the channel's Choi matrix and any bell state.
np4 Maximum "amount" of Pauli channel that can be subtracted from the input Pauli channel, such that what remains is still a valid quantum channel.

Calibrate a quantum channel

The metrics defined above can be evaluated on a set of quantum channels using the chcalib command, producing a plot as in the figure below (for the case of the Amplitude damping channel with various damping rates, cf. here).

The chcalib command can be invoked on a set of channels defined by a name and a parameter range specified in the format: <lower limit>,<upper limit>,<number of points>. If the channel has multiple parameters, thier respective ranges must be separated by ;. The noise metrics to be computed for every channel must also be specified. Multiple metric names can be provided separated by commas. If unspecified, the default metric is the infidelity (infid). For multiparamter channels, one can choose one (or two) of the noise parameters to analyze the change in the noise metrics with respect to the chosen parameter. If two parameters are chosen, the result is a 2D color plot. In summary, the usage of chcalib is:

chcalib <channel name> <noise range> [<metrics>] [<xcol>] [<ycol>].

Note that the last three parameters are optional. When chcalib is invoked, the noise metric values for all the channels are also stored in a numpy formmated file in chflow/temp/. If metric values are already present for the channels in question, then they are not recomputed and simply read from this file.

  • Physical noise processes
    • Definitions of quantum channels
    • Representations of quantum channels
    • Approximations to a Pauli channel
  • Quantum error correction
    • Quantum error correcting codes
    • Decoding and effective channel
  • Running simulations
    • On a local computer
    • On Compute Canada clusters
  • Plotting results
  • Deriving new measures of noise strength
    • Fitting logical error rates to an ansatz
    • Using machine learning techniques
Clone this wiki locally