Convolutional Networks
Simran Kaur, Rashmi Agrawal · 2021
Convolutional networks are known as convolutional neural networks when they have a grid like structure. Examples of CNNs include time series data which is seen as a 1-D grid taking samples at regular time intervals and image datasets which can be thought of as a 2-D grid of pixels. CNN is a deep neural network originally designed for image analysis. Recently, it was discovered that CNN also has an excellent capacity in consequent data analysis, such as natural language processing. CNN always contains two basic operations, namely convolution and pooling. A convolution operation using multiple filters is able to extract features (feature map) from the data set, through which their corresponding spatial information can be preserved. The pooling operation, also called sub-sampling, is used to reduce the dimensionality of feature maps from the convolution operation. Max pooling and average pooling are the most common pooling operations used in CNN. Due to the complicity of CNN, rely is the common choice for the activation function to transfer gradient in training by back propagation. In this chapter, we define what a convolution is, its need in neural networks, and the application of pooling on different datasets. This chapter also addresses how to use a CNN and the kind of operations it applies on a dataset. In this chapter, CNN are applied on image assets for feature extraction and dimensionality reduction.