In this paper we present a new scheme of a kernel-based regularization learning algorithm, in which the kernel and the regularization parameter are adaptively chosen on the base of previous experience with similar learning tasks. The construction of such a scheme is motivated by the problem of prediction of the blood glucose levels of diabetic patients. We describe how the proposed scheme can be used for this problem and report the results of the tests with real clinical data as well as comparing them with existing literature.

We propose a new CG-EGA, the PRED-EGA, for the assessment of glucose predictors. The presented analysis shows that, compared with the straightforward application of the CG-EGA, the PRED-EGA gives a significant reduction of the misclassification cases. A reduction by a factor of at least 4 was observed in the study. Moreover, the PRED-EGA is much more robust against uncertainty in the input and references.

We discuss a new regularization scheme for reconstructing the solution of a linear ill-posed operator equation from given noisy data in the Hilbert space setting. In this new scheme, the regularized approximation is decomposed into several components, which are defined by minimizing a multi-penalty functional. We show theoretically and numerically that under a proper choice of the regularization parameters, the regularized approximation exhibits the so-called compensatory property, in the sense that it performs similar to the best of the single-penalty regularization with the same penalizing operator.

Abstract. In this paper, we consider the problem of estimating the derivative of a function from its noisy version contaminated by a stochastic white noise and argue that in certain relevant cases the reconstruction of by the derivatives of the partial sums of Fourier–Legendre series of has advantage over some standard approaches. One of the interesting observations made in the paper is that in a Hilbert scale generated by the system of Legendre polynomials the stochastic white noise does not increase, as it might be expected, the loss of accuracy compared to the deterministic noise of the same intensity. We discuss the accuracy of the considered method in the spaces L2 and C and provide a guideline for an adaptive choice of the number of terms in differentiated partial sums (note that this number is playing the role of a regularization parameter). Moreover, we discuss the relation of the considered numerical differentiation scheme with the well-known Savitzky–Golay derivative filters, as well as possible applications in diabetes technology.