default_analysis_options.ini

[DEFAULT]
run = True

[Meta Data]

[Ablation]

[Algorithm Footprint]

[Auto-PyTorch Overview]

[Auto-PyTorch Tensorboard]

[Best Configuration]

[Budget Correlation]

[BOHB Learning Curves]

[Configurator Footprint]
# whether or not to have a time_slider-widget on cfp-plot
# INCREASES FILE-SIZE DRAMATICALLY
time_slider = True
# maximum number of data-points to plot
max_configurations_to_plot = -1
# if use_timeslider is not off, defines the number of quantiles for the
# slider/ number of static pictures
number_quantiles = 10
# whether to use a logarithmic scale for the timeslider/quantiles
timeslider_log = True

[Cost Over Time]
# from ['racing', 'minimum', 'prefer_higher_budget'], defines incumbent trajectory from hpbandster result
incumbent_trajectory = racing
# if True, average over plots. if False, all runs are treated individually with checkboxes
average_over_runs = True

[empirical Cumulative Distribution Function (eCDF)]

[fANOVA]
fanova_pairwise = True
pimp_max_samples = -1

[Feature Clustering]

[Feature Correlation]

[Feature Importance]

[Forward Selection]

[Importance Table]
sort_table_by = average

[Incumbents Over Budgets]

[Local Parameter Importance (LPI)]

[Parameter Importance]
whisker_quantiles_plot = True
interactive_bokeh_plots = True

[Parallel Coordinates]
params = 5
pc_sort_by = all
n_configs = 100
max_runs_epm = 300000

[Performance Table]

[Scatter Plot]

[Violin and Box Plots]