Key points are not available for this paper at this time.
Automation of data processing processes is an important direction in the field of information technology. The main focus of researchers is usually on training intelligent systems. One of the key aspects of this process is the selection of hyperparameters for models. This research paper analyzes a combined method for tuning hyperparameters in a classification mathematical model. The method integrates the functionalities of two well-established approaches: exhaustive search and limited search. Initially, the first approach is employed to discover a preliminary estimation of the model’s quality metric’s maximum value. Subsequently, the second approach is utilized to generate a final estimation of achievable quality and compile a list of hyperparameter value combinations that optimize the classifier’s efficiency. To verify the validity of the method, custom software was developed using the stochastic gradient descent algorithm. The results obtained from the experiment demonstrate the effectiveness of the proposed method.
A Mon, study studied this question.