Labeled optimal partitioning

In data sequences measured over space or time, an important problem is accurate detection of abrupt changes. partially labeled data, it to correctly predict presence/absence changes in positive/negative regions, both the train and test sets. One existing dynamic programming algorithm designed for prediction unlabeled regions (and ignores labels set); another fitting (but does not changepoints regions). We resolve these issues by proposing a new optimal changepoint model that guaranteed fit can also provide predictions data. propose algorithm, Labeled Optimal Partitioning, we formal proof solves resulting non-convex optimization problem. theoretical empirical analysis time complexity our terms number size sequence segment. Finally, evidence more than baselines, label error.