Neural Networks and Deep Learning

An artificial neural network is one of the most performant model architectures of machine learning, inspired by the biological neural network, which describes the way animal and human brains process external inputs. Such networks learn how to perform tasks by processing information, even without being programmed with task-specific rules. Artificial Neural Networks have received a lot of interest in machine learning research and industrial applications due to their flexibility and performance. An Artificial Neural Network is composed of a collection of connected units or nodes, called artificial neurons. Each connection within the network, called edge or link, allows the transmission of information to other neurons as synapses in the animal brain. A neuron receives the information, does the processing, and transmits the elaborated information to other neurons. Neurons and edges have weights determining the intensity of the information transferred at each connection. Typically, information passes only after a certain threshold. Series of neurons forms layers, where different layers may process information in many ways. The information passes from the first layer to the last one, eventually passing through each layer multiple times. This chapter presents different neural network architectures, including Feedforward and Convolution Neural Networks.