2'''. Prune the descriptors from Bag of features and AlexNet

Performed by: PruneLearners.m

This code prunes the descriptors based on maximising the variance across classes as described in the paper. This allows to identify the descriptors that are important for classification. It is typically done on manually annotated dataset.

Input: This requires annotated (manually typically) descriptor file.

Paramaters: The number of descriptors to be kept post-pruning can be set (here 6).

Output: The code saves the descriptors with the best variance.