Abstract

Neural networks (also called artificial neural networks) refers to a class of machine learning methods that was developed in multiple fields including statistics and artificial intelligence. This entry discusses several topics with regard to neural networks in three sections. The first section introduces the model and theories of the quintessential neural network model, feed-forward neural networks, under the framework of supervised learning. It focuses on the training and regularization of neural networks models given their uniqueness among other machine leaning models. The second section discusses the modern developments of neural networks that lead to its successful application in the various domains of artificial intelligence. Such developments are grouped into three categories: architecture design, methodology refinements, and advances in software and hardware. The third section briefly illustrates neural networks applied in the other two major paradigms of machine learning: unsupervised learning and reinforcement learning.