|
|
 Выпуск 89
- Zhilyakova L., Chaplinskaya N. Research of complete homogeneous "greedy-vertices" resource networks
The modification of the graph dynamic model "resource network" – "resource network with greedy vertices" is described. In this model, the graph vertices at each discrete-time moment exchange resources through the edges with limited throughput, first passing the available resource to themselves via the loop and then distributing the remaining resource to adjacent vertices according to the "standard" resource network rules. These are two rules with threshold switching: if the vertex resource exceeds the total throughput of all its outgoing edges, it gives away the full throughput to each edge; otherwise, it gives the entire available resource, distributing it in proportion to the throughputs of the outgoing edges. The process of functioning of a complete homogeneous resource network with "greedy" vertices at different values of total resource and different initial states is investigated. Possible network states are described; a non-standard state – the shutdown of the network has been identified. Two total resource thresholds, separating zones of different network behavior, have been found: the first threshold divides the zones of "insufficient" and "sufficient" resources, the second divides the zones of "sufficient small" and "sufficient large" resources. For each zone the functioning of the network is described and the asymptotic states and flows are investigated. The examples demonstrating numerical experiments are given for all typical situations.
- Kalmykov N., Sidelnikov Y. Experimental studies on the test questions based on the point ratio
The article proposes the existence of an objective component of the complexity of the problem which is experimentally verified and proposes the point ratio which is introduced and experimentally studied. For this purpose, the so-called "point coefficient" is introduced and experimentally investigated, which allows us to estimate the level of complexity of the problem given in the form of a test questions. This coefficient is based on the answers to these questions that the level of complexity of the problem is estimated. Hypotheses describing the dynamics of changes in the values of the point coefficient are proposed and experimentally tested. Also, the stability of the obtained conclusions based on the results of calculations for two qualitatively different coefficients relative to the number of respondents was checked. The criterion for determining the test question as a problem of the third level of complexity is experimentally found and tested. The reason for this is the nature of the change in the range of the spread of the values of the histogram coefficient. An experimental study of the presence of cognitive distortions in the assessment of some test tasks by experts was conducted. Possible directions for further development are proposed.
- Ponomarev A. Application of game-theoretic methods and models to the design of large-scale human-machine computing systems
Systems using elements of large-scale human-machine information processing (crowdsourcing, crowd computing) have recently become popular in solving a number of practical problems. One of the main problems associated with the use of human-machine computing is the quality of the results obtained in this way. A promising approach to the design of such systems is the use of game-theoretic modeling of the situation (system participant and method of reward distribution), to ensure that the method of reward distribution contributes to the participant applying maximum efforts and, accordingly, achieving the goals of the system designer. The paper presents the results of a review in the field of application of game-theoretic models for the rational design of systems employing large-scale human-machine computing. In particular, assumptions used in game-theoretic modeling of such systems and the main classes of models are identified. The obtained results are grouped in accordance with the typology of large-scale human-machine computing systems; the paper shows what assumptions, modeling goals, and types of models are typical for each distinguished type of such systems. The review may be useful both to practitioners who are constructing software systems that include elements of large-scale human-machine computing, and to researchers working in this field.
- Obukhov D. Speech recognition system for russian-language telephone speech
We describe a system designed to recognize Russian-language speech. Our focus is on the domain of telephone conversations, when a single-channel noisy audio signal with a sample rate of 8 kHz is received at the input. Additionally, data from YouTube video hosting is used for training. We consider a number of acoustic models and techniques for building a lexicon and language model. In addition, we conduct experiments on the influence of speaker information. It is also shown that the use of augmentation techniques such as reverb, changing the speed and volume of a signal, masking frequency and time characteristics significantly increase the quality of recognition. We achieve word error rate 24.21 on our validation dataset.
|
|
