Assumption of same variance in Linear Discriminant Analysis?




In LDA for dimensionality reduction, there is an assumption that the data points for each class come from a Gaussian distribution with same variance,but different means.

Can somebody please help me regarding the reason behind making this assumption.?


Like all assumptions in mathematics this one is also made to make your life easier. Since you don’t know the distribution of points in each class you assume it to come from Gaussian Distribution so that it is easier for you to work on the data set.