Closed Datseris closed 11 months ago
As far as I can see, Changepoints.jl
recognizes the changepoint when it occurs and not before. Therefore it would be a very valuable tool to recognize changepoints in e.g. complexity measures.
I propose we implement indicators that would show such jumps and then consider using Changepoints.jl
within TransitionIndicators.jl
to recognize them.
If we use exactly our framework, as it is now, and then use as the indicator metric eg. complexity entropy, and as the change metric e.g., any distribution distance function, we also find change points where it occurs. It is just a matter of what is the indicator and what is the change metric. ChangePoints.jl however uses a compelte different algorithm to find a change point. We use significance testing via surrogates to find a change point. They use optimizing a cost function.
So the question is, how to unite these two compeltely different approaches under a common API.
Closing in favor of #52
And we have solved it: we now have a way to unite these completely different approaches under a common API.
opening this quickly before I forget, and will write more later. Software with similar goals exists and we have to think how to integrate or incorporate to not duplicate effort or content: https://github.com/STOR-i/Changepoints.jl