Any cases where Dev set error > Train set error?

Course Q&A Deep Learning Specialization Improving Deep Neural Networks: Hyperparameter tun
1

Hi!
I had a silly doubt; are there any possible cases where the Variance is negative? Is it possible that the dev set error is lower than the train set error?

2

Bias and variance are measurements of model performance, a negative value doesnt make any sense. Yes the model can perform better in dev set if the dev set is an “easy” subset compared to train set.