Introduction to automated machine training (Automl)

The stunning success of commercial machine training applications (Machine Learning - ML) and the rapid growth of this industry created a high demand for ready -made ML methods that can be easily used without special knowledge. However, today, the success of practical application is decisive depends on experts - people who manually choose suitable architectures and their hyperparameters. Automl methods are aimed at eliminating this narrow place by constructing ML systems capable of automatic optimization and self -support, regardless of the type of input data.

In this book, a comprehensive review of the basic methods of automated machine learning (Automl) is presented for the first time. The publication will serve as a starting point for studying this rapidly developing area, for those who already use Automl in their work, the book will come in handy as a reference book.

Among the topics under consideration:
- optimization of hyperparameters,
- teaching the model based on the properties of the task,
- a review of methods for NAS,
- systems and frameworks and frameworks Automl,
- the results of the first competitions in Automl,
- problems of automated machine learning