what is the role of gamma and epsilon in the calculation of K-L divergence?

Question

I was wondering if anyone can explain to me why gamma and epsilon might be used to calculate the KL divergence between two documents? What are their roles? I am not really good with maths so if someone can explain to me in simple terms that would be very helpful.

Thank you for the help!

Answer 1

I suppose you are referring to the gamma and epsilon values defined in the paper Using Kullback-Leibler Distance for Text Categorization .

epsilon is the probability of a term which is not in a document. It is set to a small value instead of 0 to avoid the distance to be infinite. gamma is a normalization coefficient to account of epsilon , so a probability of a term in a category satisfies the properties of a probability (sum to 1).