Domain Selection for Gaussian Process Data: An Application to Electrocardiogram Signals

Area of Science:

• Statistics
• Machine Learning
• Data Science

Background:

• Gaussian processes and Kullback-Leibler divergence are fundamental in statistics and machine learning.
• Identifying regions where probabilistic models differ is crucial for various analytical tasks.

Purpose of the Study:

• To introduce and investigate the local Kullback-Leibler divergence for pinpointing intervals of maximum difference between two Gaussian processes.
• To address the challenges in estimating local divergences and their maximum intervals.

Main Methods:

• Development of a novel method based on local Kullback-Leibler divergence.
• Utilizing Monte Carlo simulations to evaluate estimation performance and numerical efficiency.
• Application to real-world data in medical research, specifically electrocardiogram signal analysis.

Main Results:

• The proposed method effectively identifies intervals where Gaussian processes exhibit the most significant differences.
• Demonstrated robust estimation performance and computational efficiency through simulations.
• Validated the practical utility of the approach in analyzing complex biomedical signals.

Conclusions:

• The local Kullback-Leibler divergence offers a powerful tool for comparing Gaussian processes.
• The method is computationally efficient and performs well in practice.
• This approach has significant potential for applications in medical signal analysis and other fields requiring nuanced model comparison.