ABSTRACT
Neural gas is basically an artificial neural network, motivated through a self-organising map. The neural gas approach has been successfully applied to clustering, image processing and pattern recognition to extract feature values and to find and identify unique patterns for the input database. In this chapter, five subjects of images are considered—animals, buildings, clouds, flowers and vehicles—to find whether the neural gas is able to distinguish the pattern of images among them. Different types of parameters are used to extract the feature value of images using neural gas: epochs, delta, iteration, alpha0 (initial value), alphaf (final value), lambda0 (initial value) and lambdaf (final value). Maximum and minimum difference values on these parameters are used as the feature values. These values are used to distinguish the images from each other. Their observed results on the basis of feature values are represented in the form of graphs, which yield good results. Artificial neural networks (ANNs) and deep learning are the main techniques for gathering the results never expected before, allowing us to become aware of a host of use cases that span research, drug development, diagnosis, treatment, prevention, patient safety and beyond. The uses are primarily in voice control for devices such as phones, tablets, TVs and hands-free speakers, enabling speech-based and other natural language inputs and structure data interactions. Training neural networks suggests an immense amount of data, and this is only for getting some feature values. Be aware that not all applications are well-suited to ANNs or deep learning, so it is important to understand when to leverage ANNs and deep learning, versus other types of machine learning techniques to achieve the desired results.
