There are two different dataset problems solved -

Hand Digit Recognition 5,6

Attributes Col - 64

Train rows - 777

Test rows - 333

Assuming data from each class can be modelled using multivariate normal distribution, then learning Parameters using Maximum Likelihood and estimating misclassification rates.

Further, altering CoVariance matrix for analysis.

Attributes - 2

Train rows - 310

Test rows - 90

Using same technique of Binary Classifier and Varying Covariance and visualising through plots of Discriminant function and isoprobability contour.