Skip to content

Iterration parametrs #1

Description

@sowhatagain

Good day.

I was using _main_make_predictions_for_second_usmacro_data_bvarsv.py with default parametrs (iter1 = 3000, iter2 =500) and i see that trend_loss ~170 (f=0) and trend/observation look like this.

Image

but if i increase iteration parametrs (iter1 = 20000, iter2 = 5000) than trend_loss ~5 and and trend/observation look like this

Image

So it looks like default params don't reach local minimum for trend_loss, and, if it reach minimum - nearly all data variation goes to trend component, basicly making VARwT part pointless.

So, please advice about iteration parametrs selection.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions