A comprehensive evaluation of heteroscedastic support vector regression via Monte-Carlo and Sieve simulation

Sunhe Wang

doi:10.1117/12.2686953

28 July 2023 A comprehensive evaluation of heteroscedastic support vector regression via Monte-Carlo and Sieve simulation

Sunhe Wang

Author Affiliations +

Proceedings Volume 12756, 3rd International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2023); 127561P (2023) https://doi.org/10.1117/12.2686953
Event: 2023 3rd International Conference on Applied Mathematics, Modelling and Intelligent Computing (CAMMIC 2023), 2023, Tangshan, China

Abstract

The heteroscedasticity in training Machine Learning (ML) methods is vital for its prediction performance. To provide practical insights about this challenge, we modify the objective function to incorporate heterogeneous variance, which enhances ML training and improves the accuracy of prediction and evaluate the performance of the proposed heteroscedastic Support Vector Regression (HSVR) method using the Sieve simulation and Monte Carlo techniques. The results demonstrate that the performance of the HSVR method continually improves as the sample size increases, with a progressively decreased root mean squared error from 0.129 to 0.028. Our findings suggest that incorporating heterogeneous variance in the objective function significantly enhances predictive performance and provides a valuable contribution to the field of ML. The proposed method can be extended to other ML models that assume a fixed variance in the noise model.

Citation Download Citation

Sunhe Wang "A comprehensive evaluation of heteroscedastic support vector regression via Monte-Carlo and Sieve simulation", Proc. SPIE 12756, 3rd International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2023), 127561P (28 July 2023); https://doi.org/10.1117/12.2686953

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available