Bridging the Gap Between Artificial Neural Networks and Kernel Regressions for Vector-Valued Problems in Microwave Applications

Thanks to their convex formulation, kernel regressions have shown an improved accuracy with respect artificial neural network (ANN) structures in regression problems where a reduced set of training samples are available. However, despite the above interesting features, inherently less flexible than ANN since implementations usually limited scalar-output problems. This article presents vector-valued (multioutput) formulation ridge (KRR) aimed at bridging gap between multioutput and scalar kernel-based approaches. The proposed KRR relies on generalized definition reproducing Hilbert space (RKHS) new structure. mathematical background is extensively discussed together different matrix functions schemes. Moreover, compression strategy based Nystrom approximation presented reduce computational complexity model training. effectiveness performance illustrative example consisting high-speed link optimization Doherty amplifier.