This repository contains all code samples described in the Machine Learning in kdb+: k-NN classification and pattern recognition with q whitepaper.
Code samples in knn_digits.q are intended to assist the user with building, testing, and benchmarking a k-NN classifier.
knn_utils.q is a usable kNN util library with the necessary functions and helpers to:
- calculate distance metric
- extract the k-nearest neighbors
- classify a test instance
- benchmark a classifier
To start the whitepaper demonstration, download the training and validation set mentioned in Chapter 1, and run from the same folder:
q knn_digits.q -s x
To load the the k-NN library:
q knn_utils.q -s x
Or
q)\l knn_utils.q