Dropout

Presentation

This optimizer is a layer optimizer that randomly disable a certain percentage of neurons during the learning phase.

Declaration

This is the function used to declare a Dropout optimizer.

OptimizerModel Dropout(float value);

Arguments

• value: The percentage of neurons to randomly disable.

Here is an example of Dropout.

StraightforwardNeuralNetwork neuralNetwork({
        Input(784),
        LocallyConnected(1, 7),
        FullyConnected(150, activation::sigmoid, Dropout(0.4f)),
        FullyConnected(70, Dropout(0.4f)),
        FullyConnected(10)
    });

Algorithms and References

The implementation of the Dropout is based on this paper :

• Dropout: A Simple Way to Prevent Neural Networks from Overfitting