Thanks for your response Murali! It is encouraging when you the see quality of content is recognized, even if it is by one person :)

I do agree that transformation like log makes more sense for calculating loss. Thanks for pointing it out and sharing the UCLA reference.