Search

Poster at FENS Conference 2022, Federation of European Neuroscience Societies Forum (https://forum.fens.org/), 9-13 July 2022, Paris, France.

Poster at Human Brain Project Summit 2023, Marseille, France, March 28-31, 2023.

Presentation at the Congress for Clinical Neuroscience with Advanced Training Academy (DGKN23) of the German Society for Clinical Neurophysiology and Functional Neuroimaging (DGKN) e.V., Hamburg, Germany, March 2-4, 2023.

Presentation at SfN 2022, San Diego, CA. Program No. 272.08.

Presentation at SWEBAGS 2022, November 30, Stockholm, Sweden. Poster 3648.

Poster at FENS Forum, 9-13 July 2022, Paris, France.

Poster presentation at SfN2022, San Diego, CA. Program No. 690.21.

The Lie-group approach was applied to determine symmetries of the third-order non-linear equation formulated for description of shear elastic disturbances in soft solids. Invariant solutions to this equation are derived and it turned out that they could represent outgoing or incoming exponentially decaying or unbounded disturbances., Comment: 5 pages

Quantization is a widely adopted technique for deep neural networks to reduce the memory and computational resources required. However, when quantized, most models would need a suitable calibration process to keep their performance intact, which requires data from the target domain, such as a fraction of the dataset used in model training and model validation (i.e. calibration dataset). In this study, we investigate the use of synthetic data as a substitute for the calibration with real data for the quantization method. We propose a data generation method based on Generative Adversarial Networks that are trained prior to the model quantization step. We compare the performance of models quantized using data generated by StyleGAN2-ADA and our pre-trained DiStyleGAN, with quantization using real data and an alternative data generation method based on fractal images. Overall, the results of our experiments demonstrate the potential of leveraging synthetic data for calibration during the quantization process. In our experiments, the percentage of accuracy degradation of the selected models was less than 0.6%, with our best performance achieved on MobileNetV2 (0.05%). The code is available at: https://github.com/ThanosM97/gsoc2022-openvino

The article discusses the technological recommendations for choosing the necessary modes of bending pipes made of corrosion-resistant steels. The article discusses in detail the engineering methodology for the construction of technological processes, a technique that includes the following steps: preparation of initial data for design; definition of the sizes of preparations and the scheme are flexible; selection of machine model; definition of bending modes; evaluation of wall thinning; adjustment of the pipe bender; bending products; control of product parameters; adjustment of the modes are flexible and sub adjustment of the machine. General conclusions were made on the work done. A feature of this technique is to determine the amount of tension, depending on the bend radius and thinning of the outer walls of the pipe and the magnitude of the pipe ovality in cross section. An important step is to control the parameters of the bending and re-adjustment of the machine in the presence of such defects as corrugation and flattening of the cross-section of the pipe.

Investigating connectivity and dynamics in the healthy and diseased striatum network using algebraic topology and data-driven in silico microcircuit reconstruction.

We compared the distribution of topological motifs, in the form of directed cliques, between different microcircuits. This technique has already provided insight on structural and functional properties of the neocortical microcircuitry. In Parkinson's Disease networks, as expected, the presence of directed cliques decreases with disease progression. Striatal interneurons despite their low percentage (5%) are crucial for the construction of high dimensional cliques.

The neuronal networks that generate locomotion are well understood in swimming animals such as the lamprey, zebrafish and tadpole. The networks controlling locomotion in tetrapods remain, however, still enigmatic with an intricate motor pattern required for the control of the entire limb during the support, lift off, and flexion phase, and most demandingly when the limb makes contact with ground again. It is clear that the inhibition that occurs between bursts in each step cycle is produced by V2b and V1 interneurons, and that a deletion of these interneurons leads to synchronous flexor–extensor bursting. The ability to generate rhythmic bursting is distributed over all segments comprising part of the central pattern generator network (CPG). It is unclear how the rhythmic bursting is generated; however, Shox2, V2a and HB9 interneurons do contribute. To deduce a possible organization of the locomotor CPG, simulations have been elaborated. The motor pattern has been simulated in considerable detail with a network composed of unit burst generators; one for each group of close synergistic muscle groups at each joint. This unit burst generator model can reproduce the complex burst pattern with a constant flexion phase and a shortened extensor phase as the speed increases. Moreover, the unit burst generator model is versatile and can generate both forward and backward locomotion.

In this work, we outline the set of problems, which any Object Detection CNN faces when its development comes to the deployment stage and propose methods to deal with such difficulties. We show that these practices allow one to get Object Detection network, which can recognize two classes: vehicles and pedestrians and achieves more than 60 frames per second inference speed on Core$^{TM}$ i5-6500 CPU. The proposed model is built on top of the popular Single Shot MultiBox Object Detection framework but with substantial improvements, which were inspired by the discovered problems. The network has just 1.96 GMAC complexity and less than 7 MB model size. It is publicly available as a part of Intel$\circledR$ OpenVINO$^{TM}$ Toolkit.

The properties of the phi-meson have been measured via its e+e- and K+K- decay channels in Au+Au collisions at sqrt{s_NN} = 200 GeV by the PHENIX experiment. The preliminary yields and temperatures derived for the minimum bias and several centrality bins in both decay channels are presented., Comment: 4 pages, 4 figures, "Quark Matter 2005" conference proceedings

First data on coherent threshold ��^0 electroproduction from the deuteron taken by the A1 Collaboration at the Mainz Microtron MAMI are presented. At a four-momentum transfer of q^2=-0.1 GeV^2/c^2 the full solid angle was covered up to a center-of-mass energy of 4 MeV above threshold. By means of a Rosenbluth separation the longitudinal threshold s wave multipole and an upper limit for the transverse threshold s wave multipole could be extracted and compared to predictions of Heavy Baryon Chiral Perturbation Theory., 7 pages, 7 figures, latex2e