reservoirpy.hyper.plot_hyperopt_report#

reservoirpy.hyper.plot_hyperopt_report(exp, params, metric='loss', loss_metric='loss', loss_behaviour='min', not_log=None, categorical=None, max_deviation=None, title=None)[source]#

Cross parameter scatter plot of hyperopt trials.

Note

Installation of Matplotlib and Seaborn packages is required to use this tool.

Parameters:

  • exp (str or Path) – Report directory storing hyperopt trials results.

  • params (list) – Parameters to plot.

  • metric (str, optional) – Metric to use as performance measure, stored in the hyperopt trials results dictionaries. May be different from loss metric. By default, ‘loss’ is used as performance metric.

  • loss_metric (str, optional) – Metric to use as an error measure, stored in the hyperopt trials results dictionaries. May be different from the default loss parameter.

  • loss_behaviour ({'min', 'max'}, optional) – How to interpret metric used as main loss in the plot. If loss need to be minimized, choose ‘min’. If loss need to be maximized, choose ‘max’. In most cases, the loss is an error function that needs to be minimized. By default, ‘min’.

  • not_log (list, optional) – List of parameters to plot with a linear scale. By default, all scales are logarithmic.

  • categorical (list, optional) – List of parameters to interpret as categorical or discrete valued.

  • max_deviation (float, optional) – Maximum standard deviation expected from the loss mean. Useful to remove extreme outliers that may create odd plots. By default, all values are kept and plotted.

  • title (str, optional) – Optional title for the figure.

Returns:

matplotlib.pyplot.figure

Matplotlib figure object.