neural_prophet icon indicating copy to clipboard operation
neural_prophet copied to clipboard

Published 20 hours ago • verified publisher icon ourownstory

Trend Regularization: Find better way to regularize

Open ourownstory opened this issue 5 years ago • 1 comments

Needed for better learning of trend deltas without gradient bleedover to older trend delta parameters. This is important if we want to achieve proper automatic trend-changepoint detection.

ourownstory avatar Jun 12 '20 21:06 ourownstory

likely not the solution but worth considering: penalize changepoints as sum of diff to previous and next segment analogous to reg term of Hodrick–Prescott_filter

ourownstory avatar Aug 10 '22 19:08 ourownstory

PR will be closed for V2 release, but ideas and feature request are saved and might be realized in V2.

leoniewgnr avatar May 18 '23 01:05 leoniewgnr