Deep learning and artificial intelligence technology make all possible: computer vision, speech recognition on mobile phones, machine translation, AI games, driverless automobiles, and other applications. We frequently interface with deep learning systems using consumer products from Google, Microsoft, Facebook, Apple, or Baidu. Computer scientist John Kelleher provides a clear, accessible, and thorough introduction to the underlying technology at the core of the artificial intelligence revolution in the MIT Press Essential Knowledge series volume.
Kelleher states that deep learning enables data-driven decisions by recognizing and extracting patterns from vast datasets; deep learning is best suited to benefit from the exponential development in big data and computer capacity due to its ability to learn from complicated data. Kelleher also provides an overview of some of the fundamental ideas in deep learning, traces its development over time, and talks about state of the art now. He discusses the most significant deep learning architectures, such as autoencoders, recurrent neural networks, and long short-term networks, as well as more recent innovations like capsule networks and generative adversarial networks. Additionally, he offers a thorough (and understandable) introduction to the two primary deep learning methods, gradient descent, and backpropagation. Kelleher then discusses the future of deep learning, including important difficulties, notable trends, and potential developments.