Data Brew Season 2 Episode 4: Hyperparameter and Neural Architecture Search
Data Brew by Databricks
00:00
The Key Idea Behind Hyperband Is Early Stopping
hyperband is distinguished by its use of early stopping. If a particular setting is not doing well, then we should instead focus on training the hyper pramber settings that appear to be more promising. I really like this idea of don't waste all your comput if you know that that model 's not going to perform well. Just cut your losses early and instead double down n the models that are performing well. And so it's just such a simple and elegt, simple and elegant idea.
Transcript
Play full episode
