Space Sampling Techniques Comparison for a Synthetic Low-Pass Filter Bayesian Neural Network

This paper presents a comparative analysis of three sampling techniques for generating space points to develop a Bayesian neural network (BNN) surrogate model of a synthetic second-order low-pass filter. The objective is to assess the effectiveness and efficiency of different sampling methods in the performance of the Bayesian surrogate model. The study draws inspiration from widely used sampling techniques such as uniform distribution, uniformly distributed random numbers, and Latin Hypercube Sampling (LHS). The results reveal that the BNN surrogate model achieves the best performance when using the LHS sampling method highlighting the impact of sampling techniques on the surrogate model's performance.