We have 4 equal classes with characteristic vectors in 2 dimensions that follow for each class 2-D Gaussian distributions with average values given and respectively with a common diagonal register of covariance which is also given in the form of a table.
In our project we were asked to create a set of training N points that stem from the above 4 classes and distributions. Using Matlab mvnrnd with appropriate parameters. It is also useful to initialize through randn ('seed', 0), so that it is possible to repeat the experiment and reproduce the results.
Repeat the above for classes with a common but non-diagonal covariance register. Commentary on the differences between the Euclidean distance classifiers, the Mahalanobis distance and the Bayesian classifier in relation to Part A.
Repeat the above for non-equal classes and for earlier classes probabilities.
return to part B and its components. In this part of the job you will assume that you have two classes that result from the union of the previous classes and you will sort the data using their projection in an appropriate straight line, based on the PCA and LDA transformations.
Hint: More details about the work can be found in the attached pdf (greek language).