Abstract: This paper proposes a scalable, matrix-free approach to kernel-based regularization for finite impulse response estimation. Our methodology is based on Bayesian optimization, a gradient-free ...