Details
-
Task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
function signature knnImputer(n_neighbors=5, weights='uniform', metric='euclidean')
The main task is to write an efficient implementation that reduces the complexity of pairwise similarity via techniques like cluster, blocking, etc.
KNN requires to compute the similarity/distance between all pairs but this pairwise similarity computation is expensive so one way to reduce this complexity is to apply the clustering in the data and then only compare the pairwise similarity inside cluster another way could be to apply meaningful blocking techniques.