K Nearest Neighbors (KNN)

Page created in the context of the course ECE662.

K nearest neighbor (KNN) classifiers do not use any model to fit the data and only based on memory. The KNN uses neighborhood classification as the predication value of the new query. It has advantages - nonparametric architecture, simple and powerful, requires no traning time, but it also has disadvantage - memory intensive, classification and estimation are slow.

Related Rhea pages:

• A tutorial written by an ECE662 student.
• ECE662 lecture notes, Spring 2008, Prof. Boutin:
  • Lecture 16: Parzen Windows and KNN density estimates
  • Lecture 17: Nearest neighbor classification
  • Lecture 18: Nearest neighbor classification and metrics
  • Lecture 19: Nearest neighbor error rate

Other reference:

• A KNN Tutorial website. Contents described below

• How K-Nearest Neighbor (KNN) Algorithm works?
• Numerical Example (hand computation)
• KNN for Smoothing and Prediction
• How do we use the spreadsheet for KNN?
• Strength and Weakness of K-Nearest Neighbor Algorithm
• Resources for K Nearest Neighbors Algorithm

• KNN
• Class Prediction using KNN
• WIKIPEDIA

Back to ECE662