2

Really simple question, when using keras-tuner and searching for the best set of hyperparameters there are a range of types to search for, so for simplicity let's say I'm using hp.Int(), now I can set a minimum, a maximum and a default value for this. My question is where does default value come into play? If we're searching for the best possible combination, why a default value (I know the Bayesian search does not actually try all possible combinations so I have a feeling it can come into play here).

DPM
  • 845
  • 7
  • 33
  • Please check [this](https://keras.io/api/keras_tuner/hyperparameters/#int-method) reference for detailed understanding in this. –  May 11 '22 at 12:32
  • Where is the detailed understanding? I checked this page before and it only states what it means, but not how it's used. What I'm asking is: when searching for the bests et of hyperparameters, where does a default value come into play when you are searching through several values of units for instance – DPM May 11 '22 at 12:43
  • 1
    Okay, `Default` value means base value or min_value where it starts searching for the best hyperparameters. If `Default` is None or not defined, it will consider min_value as the default value. –  May 11 '22 at 15:16
  • So when you specify both, when is the default value used instead of the min value? – DPM May 19 '22 at 14:30

0 Answers0