An Overview of Selected Autoencoders and Their Potential Application in Smart Cities

Abstract

The following work gives an overview of a special type of neural networks, autoencoders, that can be of great interest to researchers and practitioners in the field of smart city, due to their numerous application possibilities in this context. Given the fact that these networks can be trained in an unsupervised fashion, autoencoders are immediately applicable to practically collected data sets that often lack labels, not requiring the tedious process of data labeling. In addition to the classical autoencoder, we present two other types, and highlight their differences in architecture and in areas of application. In doing so, the benefits of the respective autoencoders and their possible application, especially in the context of smart cities, are presented.