where Pprune ={ τ ˆTk,n 1 |Lk,n 1(z,τ)> L

For each τ ˆTk,n, the corresponding loss function atτ is written as a positive definite quadratic function. Additional pruning condition for parametric DP when K is fixed. In 3.3, we showed that BinSeg-SI hadbeen considered asacomputationally efficient approximation ofthe problem in (7), where the authors additionally condition on extra information for computational tractability, e.g., the order that CPs are detected. This is one of the reasons why BinSeg-SI has lowpower. In SI, there are mainly two approaches to characterize the selection event.