Error Multiplier

Presentation

This optimizer just multiply the error of the layer by a value to increase the learning of this layer. It is mainly a test optimizer, avoid using it.

Declaration

This is the function used to declare a Error Multiplier optimizer.

OptimizerModel ErrorMultiplier(float factor);

Arguments

factor: The factor by which the error will be multiplied.

Here is an example of ErrorMultiplier.

 StraightforwardNeuralNetwork neuralNetwork({
        Input(32, 32, 3),
        Convolution(2, 3, activation::GELU, ErrorMultiplier(50.0f)),
        FullyConnected(100, activation::sigmoid, Dropout(0.2f)),
        FullyConnected(200, activation::sigmoid, Dropout(0.2f)),
        FullyConnected(10)
        },
        StochasticGradientDescent(0.003f, 0.2f));