This paper is a survey paper that summarizes hyperparameter optimization methods and related technologies in machine learning models.

It is of great value to researchers and students who are about to start research in this rapidly developing field, corporate engineers who develop machine learning systems, and researchers in different fields who are considering the introduction of machine learning. It is also extremely valuable as a survey in the field that can be read in Japanese, and this paper is considered a necessity for beginners.

This survey paper summarizes the requirements for hyperparameter optimization methods and carefully explains typical black box optimization methods. After that, this paper introduced a gray box optimization method aimed at speeding up hyperparameter optimization.

Finally, the guidelines for algorithm selection according to the situation are summarized. The guidelines for algorithm selection are a particularly important topic, and this survey paper introduces a large number of documents, and this topic is comprehensively covered. In the machine learning model, it is necessary to set appropriate hyperparameters.

It has been reported that even the classical method can exhibit better performance than recent deep learning models by setting appropriate hyperparameters, and the importance of chewing is increasing.

This survey paper explains the optimization of hyperparameters for practical use, and not only explains each method, but also gives guidance on which method should be used. Therefore, it can be said that this paper is useful for many readers who are actually thinking about application.

It is highly commendable that this survey paper not only introduces various methods, but also describes the basic knowledge, and finally the guidelines for deciding which method to use.

The explanation is concisely, and the quality and quantity are organized in an easy-to-understand manner for beginners, and this survey paper is expected to have a significant ripple effect on the introduction of machine learning.