This paper deals with the application of the support vector machine (SVM) and the least-squares SVM regressions to the uncertainty quantification of complex systems with a high-dimensional parameter space. The above regression techniques are used to build accurate and compact surrogate models of the system responses from a limited set of training samples. The accuracy and the feasibility of the proposed modeling techniques are then investigated by comparing their results with the ones predicted by a sparse polynomial chaos expansion by considering two real-life problems with 8 and 30 random variables, respectively.

This paper investigates the application of support vector machine to the modeling of high-speed interconnects with largely varying and/or highly uncertain design parameters. The proposed method relies on a robust and well-established mathematical framework, yielding accurate surrogates of complex dynamical systems. An identification procedure based on the observation of a small set of system responses allows generating compact parametric relations, which can be used for design optimization and/or stochastic analysis. The feasibility and strength of the method are demonstrated based on a benchmark function and on the statistical assessment of a realistic printed circuit board interconnect, highlighting the main features and benefits of this technique over state-of-the-art solutions. Emphasis is given to the effects of the initial sample size and of input noise on the model estimation.

Simultaneous PET/MR/EEG (Positron Emission Tomography – Magnetic Resonance – Electroencephalography), a new tool for the investigation of neuronal networks in the human brain, is presented here within the framework of the European Union Project TRIMAGE. The trimodal, cost-effective PET/MR/EEG imaging tool makes use of cutting edge technology both in PET and in MR fields. A novel type of magnet (1.5T, non-cryogenic) has been built together with a PET scanner that makes use of the most advanced photodetectors (i.e., SiPM matrices), scintillators matrices (LYSO) and digital electronics. The combined PET/MR/EEG system is dedicated to brain imaging and has an inner diameter of 260 mm and an axial Field-of-View of 160 mm.It enables the acquisition and assessment of molecular metabolic information with high spatial and temporal resolution in a given brain simultaneously. The dopaminergic system and the glutamatergic system in schizophrenic patients are investigated via PET, the same physiological/pathophysiological conditions with regard to functional connectivity, via fMRI, and its electrophysiological signature via EEG. In addition to basic neuroscience questions addressing neurovascular-metabolic coupling, this new methodology lays the foundation for individual physiological and pathological fingerprints for a wide research field addressing healthy aging, gender effects, plasticity and different psychiatric and neurological diseases.The preliminary performances of two components of the imaging tool (PET and MR) are discussed. Initial results of the search of possible candidates for suitable schizophrenia biomarkers are also presented as obtained with PET/MR systems available to the collaboration.

This paper develops a fully automatic tool for the denoising of partial discharge (PD) signals occurring in electrical power networks and recorded in on-site measurements. The proposed method is based on the spectral decomposition of the PD measured signal via the joint application of the short-time Fourier transform and the singular value decomposition. The estimated noiseless signal is reconstructed via a clever selection of the dominant contributions, which allows us to filter out the different spurious components, including the white noise and the discrete spectrum noise. The method offers a viable solution which can be easily integrated within the measurement apparatus, with unavoidable beneficial effects in the detection of important parameters of the signal for PD localization. The performance of the proposed tool is first demonstrated on a synthetic test signal and then it is applied to real measured data. A cross comparison of the proposed method and other state-of-the-art alternatives is included in the study.

This article introduces a probabilistic machine learning framework for the uncertainty quantification (UQ) of electronic circuits based on the Gaussian process regression (GPR). As opposed to classical surrogate modeling techniques, GPR inherently provides information on the model uncertainty. The main contribution of this work is twofold. First, it describes how, in a UQ scenario, the model uncertainty can be combined with the uncertainty of the input design parameters to provide confidence bounds for the statistical estimates of the system outputs, such as moments and probability distributions. These confidence bounds allow assessing the accuracy of the predicted statistics. Second, in order to deal with dynamic multioutput systems, principal component analysis (PCA) is effectively employed to compress the time-dependent output variables into a smaller set of components, for which the training of individual GPR models becomes feasible. The uncertainty on the principal components is then propagated back to the original output variables. Several application examples, ranging from a trivial RLC circuit to real-life designs, are used to illustrate and validate the advocated approach.

This article presents a new noniterative inverse modeling technique based on machine learning regression and its applications to microwave design optimization. The proposed inverse model accepts the high-dimensional S-parameters computed at many frequency points as the input and estimates the optimal geometrical/physical parameters of the microwave component as its output. The least-squares support vector machine regression is combined with the principal component analysis to simultaneously overcome both the high-dimensional input space and ill-posed challenges of the inverse modeling. We also propose a new empirical method to find the optimum number of principal components (i.e., the compression level) for each example in an automated way. This makes our proposed model general and easy to use compared with the existing data-driven inverse modeling techniques. The inverse model is trained by a set of scattering parameters computed via a 2-D/3-D solver for few configurations of the geometrical parameters. The feasibility and the accuracy of the proposed optimization scheme are investigated by comparing its predictions with the corresponding optimal configuration estimated via a commercial solver.