Pelt convergence for large penalties

[JB-Git-15]

Hi
I'm trying Pelt, and I see that the computation time increases a lot when using large penalties.
Is it possible to modify some tolerance parameter or the number of iterations ?
Best regards

JB

[deepcharles]

Hi, if you use rpt.Pelt, you can change the jump parameter as follows: algo = rpt.Pelt(jump=k). It means that only multiple of k are considered when looking for change point indexes (k, 2k, 3k, etc.). By default, k is set to 5 and if you increase it (to, say, 10) the algorithm should be faster.