Release date: November 2017

Deep learning is a type of machine learning that empowers computers to learn from experience. The book contains the mathematical and conceptual foundations of linear algebra, probability and information theory, numerical computation, and machine learning to the extent necessary to understand the material. Deep learning techniques used in practice are described, including direct propagation deep networks, regularization, optimization algorithms, convolutional networks, sequence modeling, etc. Applications such as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and video games are discussed. Finally, promising research directions are described: linear factor models, autoencoders, representation learning, structural probabilistic models, Monte Carlo methods, statistical summation, approximate inference, and deep generating models. The publication will be useful to undergraduate and graduate students as well as experienced programmers who would like to apply deep learning as part of their products or platforms.

Briefly about the topics:

• basic mathematical apparatus;
• well-known deep learning algorithms;
• linear algebra, probability theory, and fundamental machine learning concepts;
• basics of machine learning;
• deep direct propagation networks;
• regularization in deep learning;
• Optimization in deep learning;
• convolutional networks;
• sequence modeling: recurrent and recursive networks;
• algorithm selection for a particular application, and collection and analysis of data obtained in order to improve the machine learning system.

About the authors:

Ian Goodfellow is a researcher at OpenAI. He has invented various machine learning algorithms, including adversarial networks, and has contributed to various machine learning programs, including TensorFlow and Theano.

Yoshua Bengio, professor in the Department of Computer Science and Operations Research and director of the Montreal Institute for Learning Algorithms. The main goal of his research is to understand the principles of learning that give rise to intelligence. He teaches a course on machine learning and supervises a large group of undergraduate and graduate students. He is editor of the Journal of Machine Learning Research and deputy editor of Neural Computation. Has co-organized a variety of conferences, seminars and symposia on machine learning.

Aaron Courville, Associate Professor in the Department of Computer Science and Operations Research at the University of Montreal and member of the Montreal Institute for Learning Algorithms (MILA).