plot — Visualize Results¶

The plot subcommand reads .parquet snapshot files (produced by snap) and generates publication-quality plots. No heavy computation is performed — it is purely a visualization tool.

Basic Usage¶

# Plot r vs clusters and all spectra
r_analysis plot --parquet-dir snapshots/ -arc -as

# Plot everything
r_analysis plot --parquet-dir snapshots/ -a

# Filter to specific runs
r_analysis plot --parquet-dir snapshots/ -r "kmeans_BD4000" -ar -as

Input¶

Flag	Type	Default	Description
`--parquet-dir`	`str` (1+)	required	Directory(ies) containing `.parquet` files
`-r`, `--runs`	`str` (list)	all	Regex patterns to filter parquet rows by keyword

Visualization Toggles¶

Per-Run Plots¶

These generate one plot per run entry in the parquet file:

Flag	Description
`-pp`, `--plot-params`	Spectral parameter maps (\(\beta_d\), \(T_d\), \(\beta_s\))
`-pt`, `--plot-patches`	Patch/cluster assignment maps
`-ps`, `--plot-cl-spectra`	\(C_\ell^{BB}\) power spectra (one per run)
`-pc`, `--plot-cmb-recon`	CMB Q/U reconstruction maps
`-psm`, `--plot-systematic-maps`	Systematic residual maps
`-ptm`, `--plot-statistical-maps`	Statistical residual maps
`-pr`, `--plot-r-estimation`	\(r\) likelihood curve (single run)
`-ppr`, `--plot-params-residuals`	Parameter maps with residuals vs truth
`-pi`, `--plot-illustrations`	Illustration plots

Aggregate Plots (Multi-Run)¶

These overlay multiple runs on a single figure:

Flag	Description
`-as`, `--plot-all-spectra`	All \(C_\ell^{BB}\) spectra overlaid
`-ar`, `--plot-all-r-estimation`	\(r\) likelihood comparison across runs
`-ah`, `--plot-all-histograms`	Histograms of parameters across all runs

Correlation Plots¶

These plot metrics as a function of run properties (e.g., number of clusters):

Flag	Description
`-arc`, `--plot-r-vs-c`	\(r\) vs number of clusters
`-avc`, `--plot-v-vs-c`	Variance vs number of clusters
`-anlc`, `--plot-nll-vs-c`	NLL vs number of clusters
`-arv`, `--plot-r-vs-v`	\(r\) vs variance

All-in-One¶

Flag	Description
`-a`, `--plot-all`	Enable all plot types above

Group Plots¶

Group multiple runs together for aggregate comparison:

r_analysis plot --parquet-dir snapshots/ \
    -g "kmeans_BD4000" "ptep_BD64" \
    -gt "K-Means" "PTEP" \
    -ar -as

Flag	Type	Description
`-g`, `--groups`	`str` (list)	Regex patterns defining named groups
`-gt`, `--group-titles`	`str` (list)	Human-readable titles for each group
`-t`, `--title`	`str` (list)	Per-row title overrides (curve labels)

Figure Customization¶

Flag	Type	Default	Description
`-o`, `--output`	`str`	`plots/`	Output directory
`--output-format`	`str`	`png`	Output format: `png`, `pdf`, or `show`
`--font-size`	`int`	`14`	Font size for all text
`--color`	`str` (list)	auto	Custom color list for curves (cycles if fewer than runs)
`--xlim`	`float float`	auto	X-axis limits for \(r\)-estimation plots
`--r-legend-anchor`	`float float`	auto	Legend position `(x, y)` for \(r\) plots
`--s-legend-anchor`	`float float`	auto	Legend position `(x, y)` for spectra plots
`--r-figsize`	`float float`	auto	Figure size `(w, h)` in inches for \(r\) plots
`--s-figsize`	`float float`	auto	Figure size `(w, h)` in inches for spectra plots
`--r-range`	`float float`	none	\(r\) fill-between range for spectra plot
`--r-plot`	`float float`	none	Two truth \(r\) values shown as vertical lines

Examples¶

Compare K-Means configurations¶

r_analysis plot --parquet-dir snapshots/ \
    -r "kmeans" \
    -arc -ar \
    --output-format pdf \
    -o figures/

Side-by-side K-Means vs PTEP¶

r_analysis plot --parquet-dir snapshots/ \
    -g "kmeans" "ptep" \
    -gt "Adaptive K-Means" "Multi-Resolution (PTEP)" \
    -as -ar \
    --r-figsize 10 6 \
    --font-size 16

Publication-quality r likelihood¶

r_analysis plot --parquet-dir snapshots/ \
    -r "kmeans_BD4000_GAL020" \
    -ar \
    --xlim -0.01 0.02 \
    --r-plot 0.0 0.004 \
    --output-format pdf \
    --font-size 18