CODEWITHSUNDEEP
neural-networkdeep-learningkeras
elexhobby
elexhobby

Reputation: 2668

What does it mean for a dropout layer to be trainable in keras?

Keras implements dropout as a layer (just like Dense or Conv2D). Any layer has the trainable property that can be set to false if you want to freeze the weights. What does it mean for the dropout layer to be trainable, since there are no weights involved here?

Upvotes: 2

Views: 1277

Answers (1)

Daniel Möller
Daniel Möller

Reputation: 86600

When the layer has no weights, the trainable parameter is irrelevant.

In Keras, the trainable parameters just defines if the weights are going to be constant or not.

Upvotes: 3

Related Questions