Neural networks and deep learning. Training course
The book examines both classical and modern models of deep learning. The chapters of the book can be divided into three groups. - Fundamentals of neural networks. The essence of many traditional machine learning models can be understood by considering them as special cases of neural networks. The first two chapters focus on understanding the relationship between traditional machine learning and neural networks. It will be shown that the support vector machine, linear and logistic regression, singular value decomposition, matrix factorization and recommendation systems are just such special cases. Along with them, such relatively new methods of constructing features as word2vec. - Fundamental concepts of neural networks are also considered. Chapters 3 and 4 are devoted to a detailed discussion of the processes of training and regularization of neural networks. In chapters 5 and 6, radial basis function networks (RBF) and bounded Boltzmann machines are considered. - Additional questions of neural networks. Chapters 7 and 8 discuss recurrent and convolutional neural networks. Chapters 9 and 10 are devoted to more complex topics, such as deep reinforcement learning, neural Turing machines, self-organizing Kohonen maps, and generative-adversarial networks. The book is intended for senior students, researchers and practitioners. Where possible, the author pays special attention to the applied aspects of using each class of methods.
No reviews found
