Using dropout during testing

I want to test an idea of using dropout during testing to create 6-7 different predictions and than average them (a similar approach to Ensemble)
Is there a convenient way to do it in Keras?

See the answer to ‘How can I obtain the output of an intermediate layer?’ at https://keras.io/getting-started/faq/ to learn how to use K.learning_phase(). That hopefully will get you on the right track!

1 Like