What should happen to categorical crossentropy after increasing / reducing dropout?

Part 1 (2017)
5

Thank you very much for your help :). Figured out what was going on and posted the reasoning here: Bugged entropy loss calculation in Keras?