How long will it take this operator call? We should have a way to answer this from the catalog. We can
keep with the (quite wrong) assumption that this depend only on the length of the time series and have a benchmark infer the time-complexity.
We could also add a second dimension (type of time series) and either have a fast way to make the correspondand for a new time series or, simpler, just use the average to better infer the time complexity.
A possible workaround to avoid a few feature computations to slow down everything would be to include timeouts. I think that would be quite complex.
How long will it take this operator call? We should have a way to answer this from the catalog. We can keep with the (quite wrong) assumption that this depend only on the length of the time series and have a benchmark infer the time-complexity.
We could also add a second dimension (type of time series) and either have a fast way to make the correspondand for a new time series or, simpler, just use the average to better infer the time complexity.
A possible workaround to avoid a few feature computations to slow down everything would be to include timeouts. I think that would be quite complex.