Top Efficiency (selections.performance.topefficiency)

Import with:

from AnalysisG.selections.performance.topefficiency import TopEfficiency

TopEfficiency

TopEfficiency is a SelectionTemplate subclass that evaluates GNN model performance for top-quark reconstruction, including efficiency, purity, invariant-mass distributions, and signal classification scores.

Output attributes:

Invariant-mass distributions (dict[str, dict[str, list[float]]]):

Attribute

Description

p_topmass / t_topmass

Predicted / truth top-quark invariant mass per decay region.

p_zmass / t_zmass

Predicted / truth Z′ invariant mass per decay region.

Probability scores:

Attribute

Description

prob_tops

dict[str, dict[str, list[float]]] — model score distribution for top-quark edges.

prob_zprime

dict[str, dict[str, list[float]]] — model score distribution for Z′ edges.

t_decay_region / p_decay_region

Truth / predicted decay-region flag distributions.

Node-count distributions:

Attribute

Description

t_nodes / p_nodes

dict[str, dict[float, int]] — truth / predicted node-count histograms.

n_tru_tops

dict[str, int] — total number of truth top quarks per region.

n_pred_tops

dict[str, dict[float, int]] — predicted top counts per score threshold.

n_perfect_tops

dict[str, dict[float, int]] — perfectly reconstructed top counts per score threshold.

Per-event score vectors (list[list[float]]):

Attribute

Description

truth_res_edge / truth_top_edge

list[int] — per-event truth resonance / top-edge labels.

truth_ntops / truth_signal

list[int] — per-event truth top multiplicity / signal flag.

pred_res_edge_score / pred_top_edge_score

Per-edge model score vectors for resonance / top classification.

pred_ntops_score / pred_signal_score

Per-event model score vectors for n-top / signal classification.