GitHub - ahmedkobtan/KNN-Classifier-for-the-MNIST-Dataset: K Nearest Neighbour (KNN) Classifier for handwritten digits in the MNIST Dataset

ahmedkobtan / KNN-Classifier-for-the-MNIST-Dataset Public

Notifications You must be signed in to change notification settings
Fork 1
Star 0

K Nearest Neighbour (KNN) Classifier for handwritten digits in the MNIST Dataset

Notifications

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
.gitattributes		.gitattributes
KNN.m		KNN.m
ReadMe.txt		ReadMe.txt
t10k-images-idx3-ubyte		t10k-images-idx3-ubyte
t10k-labels-idx1-ubyte		t10k-labels-idx1-ubyte
train-images-idx3-ubyte		train-images-idx3-ubyte
train-labels-idx1-ubyte		train-labels-idx1-ubyte

Repository files navigation

Training on 10000 training images and testing 100 testing images
- Corresponding file is KNN.m file
- The K parameter is specified before the testing loop and can be changed
- Takes almost 30 minutes to finish running, but yields very low % error compared to Linear Regression
- ‘scalars’ matrix stores distance for one testing image between all training samples
- percentage error is output in the end
- K is set to 13, since it yields only 4% error.