Laboratory of Information Technologies

Date and Time: Wednesday, 28 February 2018, at 3:00 PM

Venue: room 310, Laboratory of Information Technologies

1. Seminar topic: «Mathematical models of information processes in the academic community»

   Speaker: Taisia S. Shiniaeva (Astrakhan State University)

   Abstract:
   
   Model of activity in the network and its analysis was proposed as a result of the research. The developed model differs from the known ones in that the process of propagation of excitation is considered within the framework of the Wiener-Rosenbluth model on a scale-free network, rather than on a regular graph. A model of the temporal dynamics of the Hirsch index of an individual scientist was proposed and investigated. The proposed model correctly describes the dynamics of the Hirsch index due to the additional (in comparison with available models) accounting for changes in the scientist’s publication activity at the initial stage of the scientific career, the distribution of articles by the number of citations, the dynamics of citing of each particular article. The analysis of the relationship between the ratings of domestic scientific journals, calculated by various organizations, was carried out. The low correlation between scientometric indicators of scientific journals obtained using different databases does not allow to create a single scale for assessing the quality of scientific publications for both domestic and international publications. Program modules for the information and analytical system of Astrakhan State University have been implemented, providing convenient storage, presentation and analysis of information on the results of intellectual activity. Scientometric analysis of journal articles in which the main results of dissertational studies are published was carried out. To date, it has not been possible to answer the question whether there are objective, adequate and informative indicators for determining the signs of violations of the publication ethics in scientific articles. A methodology for assessing the effectiveness of current scientific research was proposed, based on an analysis of the prestige of journals in which the results of scientific research of the scientific collective are published. The developed indicator is focused on stimulation of employees to the publication of works in high-ranking journals in compliance with the principles of publishing ethics.
   
   

2. Seminar topic: «The classification of diffraction images by type biomolecules studied in experiments on lasers with free electrons using machine training methods»

   Speaker: Bobkov S. A. (NRC “Kurchatov Institute»)

   Abstract:
   
   The method of coherent x-ray diffraction microscopy (CXDI) on free electron lasers provides a way for determining a three-dimensional structure of nano-dimensional objects, including biological particles such as proteins and viruses, with a resolution up to 1 Å. The author has developed methods for classifying diffraction images obtained in CXDI experiments according to the type of the initial particle structure. The methods are based on machine learning approaches: support vector method, cluster analysis, artificial neural networks and others. The study of the effectiveness of the proposed methods on different sets of experimental data obtained on the free electron laser LCLS (Stanford). Investigated was accuracy and completeness of the classification, a speed of the approaches, an optimal size of the training sets in the classification. The paper studies a possibility of classifying images obtained in experiments on European XFEL in a streaming mode.
   (materials of a PhD thesis)