logoalt Hacker News

empikolast Sunday at 3:40 PM0 repliesview on HN

The number of runs you can afford are not enough to perform Bayesian optimization. Count how many different options they explored in the text and take a guess how many samples you need to start modeling the hyperparameter space.