xpectra.plotting

Publication figures for the trust study.

Colors use the validated categorical palette (light-mode steps); marks are thin, grids recessive, identity carried by legend + direct labels, never color alone.

Functions

`ci_boxplot`(groups, path)
`ci_vs_confidence`(ci, confidence, path[, rho])
`gap_barplot`(df, path)	Grouped bars: random-split vs LOSO macro-F1 per model.
`reliability_diagram`(table, title, path[, ece])
`risk_coverage_plot`(curves, title, path)
`umap_overlay`(emb_lab, class_labels, emb_unk, ...)

xpectra.plotting.reliability_diagram(table, title, path, ece=None)[source]

Parameters:

table (DataFrame)
title (str)
path (str | Path)
ece (float | None)

Return type:

xpectra.plotting.gap_barplot(df, path)[source]

Grouped bars: random-split vs LOSO macro-F1 per model.

Parameters:

df (DataFrame)
path (str | Path)

Return type:

xpectra.plotting.risk_coverage_plot(curves, title, path)[source]

Parameters:

curves (dict[str, DataFrame])
title (str)
path (str | Path)

Return type:

xpectra.plotting.umap_overlay(emb_lab, class_labels, emb_unk, path)[source]

Parameters:

emb_lab (ndarray)
class_labels (ndarray)
emb_unk (ndarray)
path (str | Path)

Return type:

xpectra.plotting.ci_vs_confidence(ci, confidence, path, rho=None)[source]

Parameters:

ci (ndarray)
confidence (ndarray)
path (str | Path)
rho (float | None)

Return type:

xpectra.plotting.ci_boxplot(groups, path)[source]

Parameters:

groups (dict[str, ndarray])
path (str | Path)

Return type: