Closed ghost closed 3 years ago
Sort of yes. If the time series have the same number of samples there is a batch method that will compute them in bulk, returning an upper or lower triangle of pair wise distances. A user reported a factor of three speedup for their problem, but in my personal experience it was only a little faster than computing the same pairs yourself (in a loop). It will really depend on the problem sizes and hardware.
This would be the _batch methods...
Thank you for providing this project.
Is there an efficient way to calculate a distance matrix for multiple time-series? This is a very useful feature for time-series analysis and is for example implemented here for dynamic time warping.