Description

Learning model complexity using max-margin interval regression. We have observed several noisy piecewise constant signals, and we have weak labels about how many change-points occur in several regions. Max margin interval regression is an algorithm that uses this information to learn a penalty function for accurate change-point detection.

Usage

Format

There are 7 related data.frames: signals contains the noisy piecewise constant signals, annotations contains the weak labels, segments and breaks contain the segmentation model, selection contains the penalty and cost information, intervals contains the target intervals of penalty values for each signal, and model describes the learned max margin interval regression model.