Evolutionary, Bayesian, and Quasi-Monte Carlo hyperparameter tuning
Loading...
Author (Corporation)
Publication date
2025
Type of student thesis
Course of study
Collections
Type
04B - Conference paper
Editor (Corporation)
Supervisor
Parent work
Information Systems for Intelligent Systems. Proceedings of ISBM 2024
Special issue
DOI of the original publication
Link
Related research data
Series
Series number
Volume
2
Issue / Number
Pages / Duration
555-567
Patent number
Publisher / Publishing institution
Springer
Place of publication / Event location
Singapore
Edition
Version
Programming language
Assignee
Practice partner / Client
Abstract
This study provides a comprehensive evaluation and comparison of different families of optimization methods, including probabilistic methods (e.g., Bayesian optimization), evolutionary, and random methods (e.g., Monte Carlo), in the context of hyperparameter tuning for a gradient boosting model applied to a multiclass classification problem. The study evaluates these methods based on their performance in finding optimal solutions, the computation time and computational resources required, and their behavior in terms of the search space exploration and exploitation trade-off. The results indicate that probabilistic methods show superior performance in exploiting optimal solutions, but at the cost of increased computational time. Monte Carlo methods, on the other hand, are based on sampling the search space, potentially leading to diverse solutions. Evolutionary algorithms demonstrate a balance between exploration and exploitation, whereas random methods show mixed behavior. The study also highlights the importance of specific hyperparameters and the complex interplay between them.
Keywords
Subject (DDC)
Event
World Conference on Information Systems for Business Management (ISBM 2024)
Exhibition start date
Exhibition end date
Conference start date
Conference end date
Date of the last check
ISBN
978-981-96-1210-9
978-981-96-1209-3
978-981-96-1209-3
ISSN
Language
English
Created during FHNW affiliation
Yes
Strategic action fields FHNW
Publication status
Published
Review
peer-reviewed
Open access category
Closed
License
Citation
Molina Van den Bosch, M., Montalbano, C., Dornberger, R., & Hanne, T. (2025). Evolutionary, Bayesian, and Quasi-Monte Carlo hyperparameter tuning. In C. So In, N. S. Londhe, N. Bhatt, & M. Kitsing (Eds.), Information Systems for Intelligent Systems. Proceedings of ISBM 2024 (Vol. 2, pp. 555–567). Springer. https://doi.org/10.1007/978-981-96-1210-9_48