Mikhail Belyaev

Learn More
We consider the problem of multidimensional approximation according to samples with a factorial experimental design (full or incomplete). Universal approximation methods do not take this peculiarity of a sample into account. In the present work, a structural approximation method is developed: the function class and the regularization are chosen in a special(More)
We describe GTApprox — a new tool for medium-scale surrogate modeling in industrial design. Compared to existing software, GTApprox brings several innovations: a few novel approximation algorithms, several advanced methods of automated model selection, novel options in the form of hints. We demonstrate the efficiency of GTApprox on a large collection of(More)
One of the key objectives of brain-computer interface (BCI) design is to construct accurate electroencephalogram (EEG) based classifier. But out of laboratory all EEG signals are contaminated with artifacts, which hamper algorithmic processing and EEG analysis, i.e. classifier ought to get a prediction for noisy data. Real-time BCI system rely on relatively(More)
This paper aims to tackle the problem of brain network classification with machine learning algorithms using spectra of networks' matrices. Two approaches are discussed: first, linear and tree-based models are trained on the vectors of sorted eigenvalues of the adjacency matrix, the Laplacian matrix and the normalized Laplacian; next, SVM classifier is(More)
In the recent years there have been a number of studies that applied deep learning algorithms to neuroimaging data. Pipelines used in those studies mostly require multiple processing steps for feature extraction, although modern advancements in deep learning for image classification can provide a powerful framework for automatic feature generation and more(More)
In this paper, we tackle a problem of predicting phenotypes from structural connectomes. We propose that normalized Laplacian spectra can capture structural properties of brain networks, and hence graph spectral distributions are useful for a task of connectome-based classification. We introduce a kernel that is based on earth mover's distance (EMD) between(More)
In this work, we study the extent to which structural connectomes and topological derivative measures are unique to individual changes within human brains. To do so, we classify structural connectome pairs from two large longitudinal datasets as either belonging to the same individual or not. Our data is comprised of 227 individuals from the Alzheimers(More)
One of the most successful Motor Imagery classification methods is the Common Spatial Pattern algorithm, which is used as a feature generation method, combined with the Linear Discriminant Analysis classifier. CSP parameters are estimated via optimizing of a criterion implicitly connected to classification accuracy. Many modifications of CSP were proposed,(More)
  • 1