arXiv subjects
Publications and source records attributed to Grigory Osipov.
In this article the generalized Lotka-Volterra model of ensemble of four excitory or inhibitory coupled elements are studied. It is shown that in the phase space of the model there exist heteroclinic network: a connected union of two or more heteroclinic cycles. A partition of the plane of coupling parameters into sets of existence of various heteroclinic networks is constructed.
We describe how some problems (interpretability,lack of object-orientedness) of modern deep networks potentiallycould be solved by adapting a biologically plausible saccadicmechanism of perception. A sketch of such a saccadic visionmodel is proposed. Proof of concept experimental results areprovided to support the proposed approach.
Using electrocardiograms as an example, we demonstrate the characteristic problems that arise when modeling one-dimensional signals containing inaccurate repeating pattern by means of standard convolutional networks. We show that these problems are systemic in nature. They are due to how convolutional networks work with composite objects, parts of which are not fixed rigidly, but have significant mobility. We also demonstrate some counterintuitive effects related to generalization in deep networks.
We propose an algorithm for electrocardiogram (ECG) segmentation using a UNet-like full-convolutional neural network. The algorithm receives an arbitrary sampling rate ECG signal as an input, and gives a list of onsets and offsets of P and T waves and QRS complexes as output. Our method of segmentation differs from others in speed, a small number of parameters and a good generalization: it is adaptive to different sampling rates and it is generalized to various types of ECG monitors. The proposed approach is superior to other state-of-the-art segmentation methods in terms of quality. In particular, F1-measures for detection of onsets and offsets of P and T waves and for QRS-complexes are at least 97.8%, 99.5%, and 99.9%, respectively.
In this study we examined the question of how error correction occurs in an ensemble of deep convolutional networks, trained for an important applied problem: segmentation of Electrocardiograms(ECG). We also explore the possibility of using the information about ensemble errors to evaluate a quality of data representation, built by the network. This possibility arises from the effect of distillation of outliers, which was demonstarted for the ensemble, described in this paper.
Phase response curve is an important tool in studies of stable self-sustained oscillations; it describes a phase shift under action of an external perturbation. We consider multistable oscillators with several stable limit cycles. Under a perturbation, transitions from one oscillating mode to another one may occur. We define phase transfer curves to describe the phase shifts at such transitions. This allows for a construction of one-dimensional maps that characterize periodically kicked multistable oscillators. We show, that these maps are good approximations of the full dynamics for large periods of forcing.
This study is focused on the mechanisms of rhythmogenesis and robustness of anti-phase bursting in half-center-oscillators (HCOs) consisting of two reciprocally inhibitory coupled neurons. There is a growing body of experimental evidence that a HCO is a universal building block for larger neural networks, including central pattern generators (CPGs) controlling a variety of locomotion behaviors in spineless animals and mammals. It remains unclear how CPGs achieve the level of robustness and stability observed in nature. There has been a vastly growing consensus in the community of neurophysiologists and computational researchers that some basic structural and functional elements are likely shared by CPGs of both invertebrate and vertebrate animals. In this study we consider several configurations of HCOs including coupled endogenous bursters, tonic spiking, and quiescent neurons, that become network bursters only when coupled by fast inhibitory synapses through the mechanism of post-inhibitory-rebound (PIR). The goal is better understanding the PIR mechanism as a key component for robust anti-phase bursting in generic HCOs