CODEWITHSUNDEEP
scikit-learnnearest-neighbor
beginner_
beginner_

Reputation: 7632

scikit-learn: Nearest Neighbors

I'm looking at sklearn.neighbors.NearestNeighbors and the corresponding BallTree class. My question is if there is any way to continue training / append new data to the tree? Or does it always have to be completely refitted?

Upvotes: 0

Views: 211

Answers (1)

afagarap
afagarap

Reputation: 649

Since Nearest Neighbors is an algorithm that checks for proximity of a given data to a cluster of data pertaining to different classes, yes, it would require refitting. There are no learned parameters (like weights in neural networks) in Nearest Neighbors, just the value of distances (e.g. Manhattan, Euclidean)

Upvotes: 1

Related Questions