Plot HVI

HyperMapper provides a method to compare different explorations on the HyperVolume Indicator (HVI) metric. Check out this link for more information on what the HVI is: https://esa.github.io/pagmo2/docs/python/tutorials/hypervolume.html

The input to the method is:

1. a json file containing the parameters of the exploration. Only the following fields are necessary in the json:

• "application_name",
• "optimization_objectives",
• "feasible_output",
• "input_parameters"

2. a list of directories each one containing one exploration with multiple possible repetitions.

The method will parse the content of the directories and collect all the csv files that are then used for the HVI computation and plot.

Example 1:

hm-plot-hvi example_scenarios/synthetic/chakong_haimes/chakong_haimes_scenario.json output_file.pdf dir1 

The pdf file will be saved in dir1.

Example 2:

hm-plot-hvi example_scenarios/synthetic/chakong_haimes/chakong_haimes_scenario.json output_file.pdf dir1 dir2 dir3

When specifying multiple directories the method plots a boxplot per directory and saves it in each directory. Some statistics on the boxplot and HVI are collected and saved in dir1_stats.txt. At the same time, we plot a lineplot file that compares all the directories' performances. This last file is saved in a location specified in the command line, i.e. output_file.pdf in the example above.