Update assessing-learning-algorithm.md

[xcg1234]

expect out-of-sample error to be greater than "our" equal to in-sample error ---> "or"