High bias, high variance

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

what should I try to solve first, high bias or high variance?

2

A first-order procedure (assuming you have enough training data):

  • High bias means your model is too simple. Add some more complex features, so you get higher variance.
  • Too-high variance means the model is over-fit. Add regularization.