Читать книгу Intelligent Connectivity - Abdulrahman Yarali - Страница 32

Deep Learning constitutes a part of the broader family of machine learning based upon the notion of artificial neural networks (ANNs). However, that is specifically a limited viewpoint of the technology at large. This has also been known for the inclusion of propositional formulae organized by multiple generative models, such as the specific nodes present in the deep belief model (deep neural networks) and deep Boltzmann machines (Chen and Zhao 2014). Across deep learning, the most apparent form of realization is the passage of data through multiple layers, wherein the data in question becomes more abstract and composite by the fold. ANNs formulate an essential aspect of this form of technology as it aims to be inspired by the biological neural networks in living beings. Particularly, these systems, when implemented, can improve themselves instead of doing some specific tasks at hand. However, one must also consider the deep neural network option, as it is an ANN but with multiple inputs and output layers. The network moves based on calculating the probability of multiple outputs and presents the seemingly most appropriate options in light of a given problem (Katsaros and Dianati 2017). Through their implementation with computer vision, speech recognition, network filtering, social media filtering, etc., deep learning has achieved a completely different domain, which has moved close to the manifestation of actual AI in terms of the improvement factor.