Open akurniawan opened 1 month ago
I think the real catch here is the DTW is a pretty expensive metric to compute, so you are going to face a bit of an uphill battle no matter what you do. One good option might be ANNchor which is an ANN approach specifically designed for dealing with expensive metrics (essentially by finding cheap approximations to the metric tailored to the training data) and would probably be a good fir for your needs.
Hi, thanks for the great package! Currently I'm trying to build a kNN with custom metrics for DTW, but the index building for just 90k of data would take around 5 hours. Do you have any suggestions to improve this?
Below is the code that I used to build the index and perform a query
Thank you!