Locally Connected layer

Presentation

This layer contains no neurons and is used to reduce the size of the input. The Max Pooling layer can be 1D or 2D depending on the previous layer.

This is the function used to declare a Max Pooling layer.

template <class ... TOptimizer>
LayerModel MaxPooling(int sizeOfPoolingMatrix)

Arguments

sizeOfLocalMatrix: The size of the matrix. For a 2D convolution the matrix is a square of length sizeOfLocalMatrix.

Here is an example of neural networks with 2D input. The Max layer receives a shape of 28 x 28 x 1 input and outputs a size of 14 x 14 x 1 output.

 StraightforwardNeuralNetwork neuralNetwork({
        Input(28, 28, 1),
        MaxPooling(2),
        FullyConnected(10)
    });