muggle14/kNNTutorial

Python notebooks for kNN Tutorial paper

kNNTutorial

Python notebooks for kNN Tutorial paper available here https://arxiv.org/abs/2004.04523:

1. kNN-Basic: Code for a basic k-NN classifier in scikit-learn.
2. kNN-Correlation: How to use correlation as the k-NN metric scikit-learn.
3. kNN-Cosine: How to use Cosine as the k-NN metric in scikit-learn. Using Cosine similarity for text classification.
4. kNN-DTW: Using the tslearn library for time-series classification using DTW.
5. kNN-Speedup: Testing the scikit-learn speedup mechanisms (kd_tree and ball_tree) on four datasets. Requires the four datasets and a py file kNNDataLoader.py to run (all available in this repo).
6. kNN-Annoy: Testing the impact of using annoy for speedup. annoy provides code for Approximate Nearest Neighbour that may not be as accurate as full k-NN. Requires kNNAnnoy.py that contains some wrapper code for annoy. Also requires the four datasets and a py file kNNDataLoader.py to run (all available in this repo).
7. kNN-PCA: Some code to use PCA to estimate the intrinsic dimension of the four datasets. Requires kNNDataLoader.py and the data files.
8. kNN-InstSel: An assessment of two instance selection algorithms (CNN and CRR) on three datasets. Requires kNNEdit.py that containst basic implementations of the two algorithms. Requires kNNDataLoader.py and the data files.