10.5 Convolutional Neural Network

Convolutional neural network (CNNs) evolved for classifying images by recognizing specific features, distinguishing each particular object class.

The network first identifies low-level features in the input image. These features are then combined to form higher-level features.

Convolutional filters determine whether a local feature is present in the image.

We can have:

• K different convolution filters in the first hidden layer
• typically apply the the ReLU activation function

Pooling Layers reduce the size of the image by a factor of two in each direction and provide some location invariance.