Waibel A. Deep Learning Neural Networks Presentation 2021

Textbook in PDF format

Deep neural network learning capitalizes on translations of basic biological constructs, such as single neuronal cells, brain regions, and cognitive networks, to their corresponding artificial intelligence counterparts, perceptrons, layers, and synthetic networks. The observed biological complexity serves as a template for designing artificial neural networks capable of learning and mimicking human intelligence. We start by examining some simple artificial neural networks for computing different Boolean operators and univariate power functions. Then, we explore more complex artificial networks designed using simpler building blocks. We will demonstrate a number of examples training artificial networks to solve supervised machine learning applications, such as predicting Titanic passengers’ survival, movie review ranking, sonar data predictions, neuroimaging and quality of life biomedical studies, handwritten digits recognition, and text and image classification. In addition, we will present strategies for transfer learning and synthetic data generation using artificial neural networks