Positive Definite Kernels in Machine Learning: Theoretical Framework (Mercer’s Theorem & RKHS) and Empirical Evaluation on Kernel Performance
DOI:
https://doi.org/10.65421/jibas.v2i4.267Keywords:
Kernel Methods, positive definite kernel, Hilbert Space, Reproducing Kernel Hilbert Space (RKHS), Radial Basis Function (RBF), Machine learningAbstract
Positive definite kernels (PDKs) play a critical role in machine learning; this paper provides a combined theoretical and numerical investigation into their behavior. We examine Linear, Polynomial, and RBF kernels, contrasting their characteristics using a benchmark dataset, a comparative analysis was performed, emphasizing the impact of each kernel on enhancing model performance.
The implementation was carried out in Python, emphasizing practical applicability and study demonstrates the positive definite kernels efficacy in handling data complexity structures. The work provides an integrated framework combining theoretical foundations with empirical validation, supporting improved model performance through informed kernel selection tailored to specific applications.

