The sample_weight is a parameter of the method fit( ) of the DBSCAN. It allows to use the benefits of the 3D without having to use the replicated points, so the speed of the algorithms is similar to the speed by using 2D. The DBSCAN didn't need modifications in order the use this new parameter, but the DDBSCAN needed.
General modifications:
I had to remove the "false cluster remotion" done by the min_samples. It is done by the length of the smallest cluster found in the DBSCAN part instead of using the min_samples now.
The parts that eliminate replicated points to send to the RANSAC or to compare to the dir_minsamples are not necessary anymore.
Specific modifications for the 2320 and 2317 runs:
I had to change the metric using by the DBSCAN to "Euclidian"
I had to enable the directional search through non-core points to find faint tracks.
The sample_weight is a parameter of the method fit( ) of the DBSCAN. It allows to use the benefits of the 3D without having to use the replicated points, so the speed of the algorithms is similar to the speed by using 2D. The DBSCAN didn't need modifications in order the use this new parameter, but the DDBSCAN needed. General modifications: I had to remove the "false cluster remotion" done by the min_samples. It is done by the length of the smallest cluster found in the DBSCAN part instead of using the min_samples now. The parts that eliminate replicated points to send to the RANSAC or to compare to the dir_minsamples are not necessary anymore. Specific modifications for the 2320 and 2317 runs: I had to change the metric using by the DBSCAN to "Euclidian" I had to enable the directional search through non-core points to find faint tracks.