Visualization

Iterative Plots

When iterative_plot is enabled during problem solving, moospread stores images of successive Pareto fronts in the directory specified by images_store_path. Several arguments of solver.solve(...) can be used to control the visualization process:

• plot_period: frequency of plotting,

• plot_dataset: whether to display the training dataset alongside the generated solutions,

• plot_population: whether to plot the full population of samples.

You may also provide a custom plotting function when initializing the SPREAD solver via the plot_func argument. The custom plotting function must accept the same arguments as solver.plot_pareto_front.

Video

A main motivation for enabling iterative_plot is to visualize the optimization process as a video after completion. This can be achieved using:

solver.create_video(
    image_folder,
    output_video,
    total_duration_s,
    first_transition_s,
    fps
)

This function creates a video from the images stored in image_folder, which are sorted according to the time index encoded in their filenames (t=...).

The first transition (from the first to the second image) lasts first_transition_s seconds, while the remaining transitions share the remaining time equally. The resulting video has a total duration of total_duration_s seconds and is rendered at fps frames per second.

In the MOBO Setting, a useful visualization is the evolution of the hypervolume over the MOBO steps. This can be obtained by setting hypervolume=True and providing the directory hv_all_value_file where the hypervolume values are stored.