Open GeorgWa opened 2 months ago
Hi, GeorgWa
Thanks for your appreciation.
For non-monotonic regression, the edges of the graph will not be subject to any monotonicity constraints, resulting in a graph with more edges. At this point, whether the longest path represents the regression curve requires further exploration. Thanks for your insightful comment.
I noticed that AlphaDIA (well done!) uses a complex multi-kernel LOESS algorithm. Perhaps it could be replaced with calib-rt.
Thanks and best, Jian Song
Hi,
This makes sense, if you have ideas how to extend it I would be quite interested. AlphaDIAs LOESS regression also uses the scikit-learn like syntax, so it should be compatible. I'm quite busy at the moment but feel free to open a PR!
Best, Georg
Hi, Really cool project!
I wondered if the requirement for monotonic increase of RT and predicted RT during graph construction precludes the application to other non-monotonic regressionn tasks like MZ and ion mobility.
Best, Georg