validation

class nomad.stop_detection.validation.AlgorithmRegistry[source]

Bases: object

Simple registry for parameterized stop-detection runs.

add_algorithm(fn, family=None, granularity=30, **param_specs)[source]

annotate_metrics(metrics, algo)[source]

family_timing_summary()[source]

iter_algorithms(family=None)[source]

record_timing(algo, elapsed_s, ping_count=nan)[source]

time_call(algo, *args, **kwargs)[source]

timing_frame()[source]

nomad.stop_detection.validation.bootstrap_metric_summary(data, metrics, group_col='algorithm', unit_col='user', n_boot=2000, interval=(0.05, 0.95), random_state=2025)[source]: Bootstrap per-unit medians and summarize them as point estimates plus intervals.

nomad.stop_detection.validation.compute_stop_detection_metrics(stops, truth, user_id=None, algorithm=None, prf_only=True, traj_cols=None, right_traj_cols=None, **kwargs)[source]

nomad.stop_detection.validation.compute_visitation_errors(overlaps, true_visits, traj_cols=None, right_traj_cols=None, **kwargs)[source]

nomad.stop_detection.validation.plot_family_timing(summary_df, ax=None, title='Mean Time Per Parameterization')[source]: Bar plot for family-level runtime summaries.

nomad.stop_detection.validation.plot_metric_boxplots(data, metrics, group_col='algorithm', group_order=None, group_families=None, colors=None, cmap='tab10', figsize=None, save_path=None, metric_titles=None, legend_title='Base Algorithm', x_tick_rotation=35, whis=(5, 95), showfliers=False)[source]: Plot per-group metric distributions as boxplots.

nomad.stop_detection.validation.plot_metric_intervals(summary_df, metrics, group_col='algorithm', group_order=None, group_families=None, colors=None, cmap='tab10', figsize=None, save_path=None, metric_titles=None, legend_title='Base Algorithm', x_tick_rotation=35)[source]: Plot precomputed point estimates with interval whiskers for each metric.

nomad.stop_detection.validation.plot_metric_vs_param(ax, data, x, y, algo_param, title='', xlabel='', ylabel='', show_max_mean_line=False, show_band=False, color='C0')[source]: Simple sensitivity plot used by validation notebooks.