Deborah.MiriamDocument

module MiriamDocument

Deborah.MiriamDocument — Document & figure generation for multi-ensemble Deborah.Miriam results.

Deborah.MiriamDocument parses a Miriam TOML config, loads reweighting/interpolation summaries across ensembles and label/train grids, computes overlap and error diagnostics, generates figures (heatmaps, reweighting curves, etc.), and saves a consolidated JLD2 snapshot.

Scope & Responsibilities

• Config parsing: read [data], [abbreviation] and related sections to resolve paths, ensemble names, model suffixes, and encoded feature/target names.
• Summary loading: aggregate RW (RWBS/RWJK/RWP1/RWP2), trace/moment measurement blocks, and cumulant blocks for all (LBP, TRP) combinations and ensembles.
• Snapshotting: write results_<overall_name>.jld2 under the analysis tree and copy a mirror to the current working directory.

Public API

• MiriamDocumentRunner.run_MiriamDocument End-to-end runner that performs loading and JLD2 emission.

Config Expectations (TOML)

• [data]: labels, trains, location, multi_ensemble, ensembles, per-target specs for $\text{Tr}\,M^{-n} \; (n=1,2,3,4)$ (*_X, *_Y, *_model, *_read_column_*, *_index_column), analysis_header, use_abbreviation.
• [abbreviation]: optional dictionary for filename $\leftrightarrow$ token encoding.

Outputs

• JLD2 snapshot <location>/<analysis_header>_<multi_ensemble>/<analysis_header>_<overall_name>/results_<overall_name>.jld2 and a copy ./results_<overall_name>.jld2.

Minimal Usage (REPL)

julia> using Deborah
julia> run_MiriamDocument("config_MiriamDocument.toml")

Notes

• Abbreviation maps, when enabled, drive compact path/name construction for learning tags.
• Column indices are $1$-based and propagated as provided in the TOML.