iqm.benchmarks.compressive_gst.mgst.additional_fns.random_seq_design

iqm.benchmarks.compressive_gst.mgst.additional_fns.random_seq_design#

iqm.benchmarks.compressive_gst.mgst.additional_fns.random_seq_design(d: int, l_min: int, l_cut: int, l_max: int, N_short: int, N_long: int) → ndarray#

Generate a set of random sequences ith given lengths.

Disclaimer: Draws without replacement but inefficiently (not working for sequence length > 24) This sequence lengths (circuit depths) are chosen for a mix of very short sequences (better convergence) and some slightly longer sequences to reduce the generalization error. This sequence design is heuristic and intended for coarse and fast estimates. For very accurate estimates at the cost of higher measurement effort it is recommended to use pyGSTi with long sequence GST.

Parameters:

  • d (int) – The number of gates in the gate set

  • l_min (int) – Minimum sequence lenght

  • l_cut (int) – Cutoff sequence lenght: N_short sequences are equally distributed among lengths l_min < l < l_cut

  • l_max (int) – N_long sequences are equally distributed among lengths l_cut + 1 < l < l_max. Currently l_max < 24 only.

  • N_short (int) – Number of short sequences

  • N_long (int) – Number of long sequences

Returns:

2D array where each row contains the gate indices of a gate sequence

Return type:

J

Contents