Deborah.RebekahMiriam

module RebekahMiriam

Deborah.RebekahMiriam — Plotting, comparison, and JLD2 I/O helpers tailored to Deborah.Miriam outputs.

Deborah.RebekahMiriam provides a Deborah.Miriam-specific companion to Deborah.Rebekah:

• loads multi-ensemble Deborah.Miriam summaries (interpolation style and measurement-at-single-$\kappa$ style),
• computes overlap/ERR dictionaries and similarity metrics (Bhattacharyya, Hellinger/JSD) against a baseline,
• renders CHK&ERR/BC/JSD heatmaps on the (LBP $\times$ TRP) grid,
• plots P1/P2 vs original along TRP at fixed LBP,
• visualizes full reweighting curves with uncertainty bands and interpolation points,
• and persists consolidated results into a single .jld2 file.

Scope & Responsibilities

• Comparison utilities Build per-observable dictionaries of overlap codes (CHK $\in$ {0,1,2}) and error ratios vs. a reference tag (e.g., RWBS), and compute Bhattacharyya/Hellinger matrices for (LBP $\times$ TRP) cells.
• Heatmaps (PyPlot.jl) Render paired CHK/ERR heatmaps and single-map BC/JSD heatmaps, with $\LaTeX$ ticks, optional annotations, contour lines, and cropped PDF export.
• PXvsBS curves Flatten Deborah.Miriam matrices to a legacy-like dict and plot P1/P2/original vs TRP at fixed LBP, for either interpolation or measurement-at-single-$\kappa$ workflows.
• Reweighting curve plot Show continuous $\kappa$-scans (RWBS/RWP1/RWP2) with bands, discrete point estimates, and final interpolated values at the chosen criterion.
• JLD2 saving Persist interpolation summaries and $\kappa$-indexed measurement summaries, plus kappa_list, rw_data, labels, trains into one .jld2.
• Summary loading Read .dat summaries into (field, :avg|:err, tag, keyword) or (… , kappa_str) matrices; derive $\kappa$ token lists from ensemble names.

Key Components

• ComparisonRebekahMiriam.build_overlap_and_error_dicts, ComparisonRebekahMiriam.build_overlap_and_error_dicts_for_measurements, ComparisonRebekahMiriam.build_bhattacharyya_dicts — construct CHK/ERR/BC grids.
• HeatmapsRebekahMiriam.render_overlap_and_error_heatmaps, HeatmapsRebekahMiriam.render_bhattacharyya_heatmap, HeatmapsRebekahMiriam.render_jsd_heatmap — CHK/ERR/BC/JSD visuals with $\LaTeX$ axes and optional contours.
• PXvsBSPlotterRebekahMiriam.plot_PX_BS_vs_trains, PXvsBSPlotterRebekahMiriam.plot_PX_BS_vs_trains_for_measurements — P1/P2 vs BS curves across TRP.
• ReweightingPlotRebekahMiriam.plot_reweighting_pyplot — $\kappa$-scan curves $+$ interpolated points (RWBS/P1/P2).
• SummaryLoaderRebekahMiriam.load_miriam_summary.
• JLD2SaverRebekahMiriam.save_miriam_results — single-file persistence.

Public API (typical entry points)

• Loading: SummaryLoaderRebekahMiriam.load_miriam_summary / SummaryLoaderRebekahMiriam.load_miriam_summary_for_measurement → (Dict, kappa_list?).
• Comparison: ComparisonRebekahMiriam.build_overlap_and_error_dicts, ComparisonRebekahMiriam.build_bhattacharyya_dicts.
• Plotting: HeatmapsRebekahMiriam.render_bhattacharyya_heatmap, PXvsBSPlotterRebekahMiriam.plot_PX_BS_vs_trains, ReweightingPlotRebekahMiriam.plot_reweighting_pyplot.
• Saving: JLD2SaverRebekahMiriam.save_miriam_results.

Minimal Usage (REPL)

julia> using Deborah
julia> using Deborah.RebekahMiriam

# Load interpolation summary and build diagnostics
julia> new_dict = SummaryLoaderRebekahMiriam.load_miriam_summary(work, ens, group, overall, labels, trains, ["skew","kurt"], [:RWBS,:RWP1,:RWP2], [:cond,:susp,:skew,:kurt])
julia> chk, err = ComparisonRebekahMiriam.build_overlap_and_error_dicts(new_dict, [:cond,:susp,:skew,:kurt], ["skew","kurt"], [:RWP1,:RWP2], :RWBS, labels, trains)  # CHK/ERR
julia> bc, _ = ComparisonRebekahMiriam.build_bhattacharyya_dicts(new_dict, [:cond], ["kurt"], [:RWP2], :RWBS, labels, trains)

# Visualize heatmaps
julia> HeatmapsRebekahMiriam.render_overlap_and_error_heatmaps(chk[(:cond,:RWP2,"kurt")], err[(:cond,:RWP2,"kurt")], N_lb_arr, N_tr_arr, :cond, :RWP2, "kurt", overall, figs_dir; save_file=true)

# P1/P2 vs BS along TRP at fixed LBP
julia> flat = PXvsBSPlotterRebekahMiriam.build_flat_plot_dict(:kurt, :RWBS, :RWP1, :RWP2, "kurt", new_dict)
julia> PXvsBSPlotterRebekahMiriam.plot_PX_BS_vs_trains("kurt", "RWBS", "RWP1", "RWP2", 25, flat, trains_int, labels_int, "kurt", overall, figs_dir; save_file=true)

# Save a consolidated JLD2
julia> JLD2SaverRebekahMiriam.save_miriam_results("out/miriam_results.jld2", legacy_summary, trace_meas, moment_meas, cumulant_meas, kappa_list, rw_data, labels, trains)

Notes

• Two workflows are supported: (i) interpolation (keyword = guiding cumulant like "kurt"), (ii) measurement-at-$\kappa$ (keyword = $\kappa$ token string like "13580"). Plotters/loaders handle both.
• Heatmaps assume integer percent ticks for LBP/TRP and can annotate cells; PDFs are auto-cropped if pdfcrop is available.
• All matrices are shaped (num_labels, num_trains) and must align with labels/trains metadata; CHK uses categorical codes, ERR is a ratio.