Skip to content

Paper figures

Each figure of the paper is one script, analysis/paper/fig_<name>.py. The published working points live in analysis/paper/utils/studies.py (reconstruction) and analysis/paper/utils/calibration.py (calibration), with the library's MODEL_RECIPE, SEED_RECIPE and FIT_RECIPE.

Redraw the published figures

The reconstructed scans and the cached inputs of every other figure are the lucid-paper-results-v1 release asset (299 MB unpacked). Install it, then plot, on a CPU:

python scripts/download_data.py --paper              # into analysis/paper/output/data/
python analysis/paper/fig_nrays.py --plot-results    # and likewise for each fig_*.py

Regenerate

Most scripts take three modes:

python analysis/paper/fig_<name>.py                   # generate the data, then plot
python analysis/paper/fig_<name>.py --generate-data   # data only
python analysis/paper/fig_<name>.py --plot-results    # plot what exists

The display figures and the calibration figures run from the configs and the example data. The three reconstruction scans (fig_nrays, fig_energy_scan, fig_geometry_scan) and the loss landscapes need 500-event PhotonSim samples that are not distributed (--root-base), and run as GPU jobs on S3DF (--backend s3df).

Every script lists its options with --help. The table of scripts, their inputs, the output layout and the S3DF campaign commands are in analysis/paper/README.md.