What is the use of dropout layers in Neural Networks
Drop out ratio = 0.2 means to drop out random 20% percent of the input features and consequently a 20% percent of the corresponding neurons in each layer.
Now each iteration has a random 80% of the features.
We do the drop out to optimize the overfitting problem.
But when we use our test data we use all the features and their corresponding neurons.
Comments
Post a Comment