When executing a ReassignTask, only certain vectors need to be checked against the entire neighborhood of clusters. Most of the vectors need to only be checked if they moved closer from before split to after split. If that is established, the information might be used in a reassign to avoid distance computations. This is in accordance with the (SPFresh paper).
When executing a
ReassignTask, only certain vectors need to be checked against the entire neighborhood of clusters. Most of the vectors need to only be checked if they moved closer from before split to after split. If that is established, the information might be used in a reassign to avoid distance computations. This is in accordance with the (SPFresh paper).