This code is the python3 implementation of a nearest neighbor classifier for hand-written digit recognition. We also include a wrapper of Scikit Learn's k-neighbor classification for comparison and validation.

To try our version of nearest neighbor classifier, simply run:

python3 mnist_dataloader.py

To try the Scikit Learn version, run:

python3 mnist_dataloader.py scikit

The resulting confusion matrix and accuracy are stored in exercise1-result.txt.

The ground truths and predicted labels are stored in label.txt and predicted.txt respectively.

#Assignment 1b

This code is the python implementation of a k-means and a k-median clustering algorithm. We also include a wrapper of Scikit Learn's k-means (in the file gonzales.py) for comparison and validation.

To try k-means, simply run:

python kmeans.py <number of clusters> method

where method can be firstk, random, kmeans++, gonz. To try the k-median , run:

python kmedians.py <number of cluster> method distance_metric 

where method can be firstk, random, kmeans++, gonz, and distance_metric can be eu or mat.

#Assignment 1c

This code is the python3 implementation of a nearest neighbor classifier for hand-written digit recognition, using Random Projections as the Dimensionality Reduction mean.

To try our version of nearest neighbor classifier, simply run:

python3 mnist_dataloader.py k

where k is the number of dimensions we want to reduce to.

The ground truths and predicted labels are stored in label_k.txt and predicted_k.txt respectively.