Any plan to implant this program to CUDA with taichi or some other lib to accelerate ?

[zhrli]

Is your feature request related to a problem? Please describe. I'm always frustrated when I deal with long data for hours.

Describe the solution you'd like To accelerate this program with GPU in Taichi or something else.

[imacocco]

Hi, which part of the pipeline are you struggling with and what is the dimensionality of the data you are dealing with?

[zhrli]

Hi, which part of the pipeline are you struggling with and what is the dimensionality of the data you are dealing with?

data are just time series with window sliding. every item is 1500 steps around.

[zhrli]

Hi, which part of the pipeline are you struggling with and what is the dimensionality of the data you are dealing with?

    for k in range(0, X.shape[0] - windowsize + 1):
        X_window = X[k:k+windowsize]
        d = MetricComparisons(coordinates=X_window, maxk=maxk)
        imb_ij, imb_ji = d.return_inf_imb_two_selected_coords([0], [1], k=1)
        imb01[0, k] = imb_ij
        imb10[0, k] = imb_ji 

I think there is a lot of parallel work to do. Cupyx gives a kdtree in GPU.