Convolutional Layer
Convolutional Layer
Convolutional layer is the core building block of a convolutional neural network. The layer’s parameters consist of a set of learnable filters (or kernels), which have a small receptive field, but extend through the full depth of the input volume. During the forward pass, each filter is convolved across the width and height of the input volume:
where is the -th feature map in the -th layer, is the -th feature map in the -th layer, is the kernel, and is the kernel size. We call the stride of the convolution the number of pixels by which we move the filter at each step.
Receptive Field
The receptive field of a neuron is the area in the input space that can affect the neuron’s output. In a convolutional neural network, the receptive field of a neuron in layer
is the area in the input space that can affect the neuron’s output in layer .
Pooling Layer
Pooling layer is a special convolutional layer used to reduce the spatial dimensions of the input volume.
Max Pooling
The most common pooling operation is max pooling, which partitions the input volume into a set of non-overlapping rectangles and, for each such sub-region, outputs the maximum value.
Proposition
If convolutional filters have size
and stride , and pooling layer has pools of size , then each unit in the pooling layer has a receptive field of size: .
Convolutional Neural Network
A convolutional neural network (CNN) is a neural network whose hidden layers contain at least one convolutional layer, and some pooling layers. CNNs are particularly effective for image recognition tasks.