CHM covariance pipeline

This folder is a cleaned public version of the Circuit Harmonic Matrix covariance workflow. It contains only the scripts needed to compute and plot Fourier-coefficient covariance/correlation data using exact two-copy parameter twirling, with optional Monte Carlo validation.

Included files

• ansatze.py — minimal PennyLane/JAX circuit library for Monte Carlo validation.
• exact_covariance_twirled.py — exact two-copy parameter-twirled covariance for one circuit configuration.
• twirled_correlation_matrices_compute.py — batch computation over qubit/depth/layer sweeps.
• twirled_correlation_matrices_plot.py — correlation heatmap grids from batch .npz outputs.
• twirled_variance_profile_plot.py — Fourier-coefficient variance profile plots.
• experiment_gui.py — Tkinter launcher for the same command-line workflow.

Not included from the uploaded working branch

• cov_estimator.py was excluded because the uploaded version is tied to parameter-tying experiments and imports tying.
• diagnose_delta_norms.py was excluded because it is a pruning/Cov-shift diagnostic and imports pruning metadata.
• Phase-alignment, Fisher/QFIM, pruning, gate-schematic, and prediction-launcher paths were removed from the GUI.

Dependencies

Core exact twirling requires numpy; plotting requires matplotlib. Monte Carlo validation additionally requires pennylane and jax. tqdm is optional for progress bars.

Example commands

Single exact covariance run:

python exact_covariance_twirled.py \
  --out_dir outputs/exact_twirled/C18_n6_L1_d4 \
  --ansatz CIRCUIT_18 --n_qubits 6 --n_layers 1 --train_depth 4 \
  --encoder_axis RX --obs_kind OZ --max_states 20000000 --progress

Batch covariance/correlation run:

python twirled_correlation_matrices_compute.py \
  --out_dir outputs/twirled_covariance/C18_n6_d1-4 \
  --ansatz CIRCUIT_18 --n_qubits_list 6 --train_depth_list 1 2 3 4 \
  --n_layers 1 --encoder_axis RX --obs_kind OZ --max_omega 6 \
  --n_theta_samples 4096 --batch_size 64 --progress

Correlation plots:

python twirled_correlation_matrices_plot.py \
  --indir outputs/twirled_covariance/C18_n6_d1-4 \
  --outdir outputs/figures/C18_n6_d1-4/correlation_heatmaps

Variance profiles:

python twirled_variance_profile_plot.py \
  --indir outputs/twirled_covariance/C18_n6_d1-4 \
  --outdir outputs/figures/C18_n6_d1-4/variance_profiles \
  --mode relmax

GUI:

python experiment_gui.py