I am trying to train a binary classifier using word vectors. I have the tfidf vectors for each sentence in my training set. Before applying binary classification algorithms, I just want to check whether the sentences belonging to two different classes, come from two different distributions. So, I basically do the following. Take the feature vector(tfidf transformed) for class 1 and class 2 and perform kernel density estimation to derive the density functions of the two. Now, I intend to do a KL divergence test to find how different are the two distributions. Is there any other way to find the difference in distributions between two feature vectors belonging to two different classes?