Review of different types of neural network architectures

Abstract

Innovative technologies come with such a huge amount of data that can only computerize with fast and more complex software. As time went by, more complicated problems arose such as pattern recognition, machine learning and prediction and unfortunately the conventional computer system was unable to carry out such tasks. Which leads to intelligent computational systems such as artificial neural networks. It is developed so that artificial neurons combined together would behave like a human brain. Different layers of mathematical processing are used to provide an accurate response regarding the input. Based on their architecture, training or learning methodology, and activation function, these artificial neurons are classified. The arrangement of neurons to create layers and the connections between and within the layers make up the neural network architecture. This paper aims to provide a clear and concise understanding of several types of architecture and its applications. Five mains' architectures and their applications and gaps are presented in this paper. The different architectures are: feed-forward, Convolutional and, recurrent neural networks, Auto encoder and generational encoders and Deep reinforcement learning architecture.