This is a project in the course Optimization and Data Analytics at Department of Engineering, Aarhus University.

For the project there are 2 different datasets:

• MNIST (Handwritten digits)
• ORL (Faces)

The MNIST data contains 10 classes, the digits from 0-9. The ORL data contains 40 classes, of 40 different people.

5 different methods have been used:

• Nearest Centroid Classifier
• Nearest sub-class centroid classifier using number of subclasses in the set {2,3,5}
• Nearest Neighbord Classifier
• Perceptron trained using Backpropagation
• Perceptron trained using MSE