Combination of recursive supervised and semisupervised filters for improved unbiased estimation

In this paper we investigate the steady-state performance of semisupervised regression models adjusted using a modified RLS-like algorithm, identifying the situations where the new algorithm is expected to outperform standard RLS. By using an adaptive combination of the supervised and semisupervised methods, the resulting adaptive filter is guaranteed to perform at least as well as the best contributing filter, therefore achieving universal performance. The analysis and behavior of the methods is illustrated through a set of examples in a plant identification setup, analyzing both steady-state and convergence situations.