Hyperparameters Tuning

Question 1

What 2 methods does he suggest for tuning?

Answer 1

1. Coarse-to-fine random searches. Which means that we randomly create a bunch of hyper parameters from a closed “grid” of options (they will have limits) test them, than zoom in the batch that work the best, set that as the limits and run again.
2. Bayesian hyper parameter optimization solutions and the code base matures