Summary: Artificial neural networks (ANNs) were designed based on the present understanding of their biological counterpart. An ANN is a system which serves as a fully parallel analog computer to mimic some aspect of cognition. Throughout the mid-2000s, many different architectures have been explored and have won contests related to machine learning and image recognition. This chapter discusses the architecture of the following types of neural networks: the perceptron model, feedforward neural networks, convolutional neural networks (CNNs), recurrent neural networks, and complex-valued neural networks. ANNs have been widely used in recent years, with applications such as image classification, speech recognition, and natural language processing. An ANN is a collection of artificial neurons constructed by connecting neurons with a weighted connection. CNNs are similar to the feedforward ANNs but are typically used to solve image and computer vision-related problems but have also been applied to natural language processing.
For the entire collection see [Zbl 1523.68010].