Warsaw, Poland

Time filter

Source Type

Slowinski G.,Vistula University | Smolinski A.,Central Mining Institute of Poland
Journal of Chemistry | Year: 2016

The continuously increasing oil prices as well as stronger environmental regulations regarding greenhouse emissions made the greatest economic powers search a new, price competitive, and environment friendly energy carrier, such as hydrogen. The world research activities in these terms focus on the development of integrated hydrogen and power generating technologies, particularly technologies of hydrogen production from various carbonaceous resources, like methane, coal, biomass, or waste, often combined with carbon dioxide capture. In the paper the thermodynamic analysis of the enhancement of hydrogen production in iron based chemical looping process is presented. In this method, iron oxide is first reduced to iron with a reducing agent, such as carbon oxide, hydrogen, or mixture of both gases (synthesis gas), and then, in the inverse reaction with steam, it is regenerated to iron oxide, and pure stream of hydrogen is produced. © 2016 Grzegorz Słowiński and Adam Smoliński.

Placzek S.,Vistula University | Adhikari B.,Vistula University
CEUR Workshop Proceedings | Year: 2014

While analyzing Artificial Neural Network structures, one usually finds that the first parameter is the number of the ANN layers. Hierarchical structure is an accepted default way to define ANN structure. This structure can be described using different methods, mathematical tools, software and/or hardware realization. In this article, we are proposing ANN decomposition into hidden and output sub networks. To build this kind of learning algorithm, information is exchanged between the first sub networks level and the second coordinator level in every iteration. Learning coefficients are tuned in every iteration. The main coordination task is to choose the coordination parameters in order to minimize both the global target function and all local target functions. In each iteration their values should decrease in asymptotic way to achieve the minimum. In article learning algorithms using forecasting of sub networks connectedness is studied.

Kowalewski O.,Warsaw School of Economics | Rybinski K.,Vistula University
Oxford Review of Economic Policy | Year: 2011

This article reviews the state's role in the transition economies of Central and Eastern Europe. Among the countries, Poland is perceived as the leader because it was the first economy to emerge from the decline following the transition, as well the only EU member state to survive the crisis without a recession. This success is often attributed to the high quality of government. However, we show that this popular perception is false and that state malfunctions on numerous fronts may soon impede Polish growth prospects. In Poland, the only bright spot is the quality of the financial supervision, which should serve as a role model for other countries. We attribute the malfunctions to the EU accession period that resulted in an unchecked growth of the government and contributed to the weakening of political and legal institutions. © The Authors 2012. Published by Oxford University Press.

Placzek S.,Vistula University
Proceedings of 2014 Science and Information Conference, SAI 2014 | Year: 2014

A Neural network with a feed-forward structure with one input, one hidden and one output layer can be presented as a hierarchical two-level structure with two independent subnetworks on both the first and the second level. This process is known as decomposition of an Artificial Neural Network (ANN) into two sub-networks. Two target functions are defined: the output target function Ψ, which defines an error function for all networks. The local target function Φ which defines the error of the first and second layer sub-network adjustment. For the coordination level, two independent functions are defined: G(V) for feed forward and H(V) for back forward. The coordinator ensures that learning algorithms for both levels, first and second, are concatenated. Solving local tasks provides for the achievement of the minimum of the global target function Ψ (global task). The article defines the obligatory conditions that have to be fulfilled (regarding both the first and the second level tasks), for the algorithm to be convergent and achieve the minimum of the global target function (the output function). A three-argument function allows us to study the general learning characteristics for both the first and the second level. Final results are discussed and the positive and negative parameters of the two stage learning algorithm are presented. Matrix weight coefficients are modified after each presentation of learning vectors X (input) and Z (output). © 2014 The Science and Information (SAI) Organization.

Zaitsev D.,Vistula University
Information Sciences | Year: 2016

When using Petri nets to investigate deadlock control, structural analysis techniques are applied, which are based on solving systems of linear algebraic equations. To gain an extra computational speed-up when solving sparse linear systems, we examine a sequential clan-composition process, represented by a weighted graph. The system decomposition into clans is represented by a weighted graph. The comparative analysis of sequential composition for subgraphs and edges (pairwise) is presented. The task of constructing a sequence of systems of lower dimension is called an optimal collapse of a weighted graph; the question whether it is NP-complete remains open. Upper and lower bounds for the collapse width, corresponding to the maximal dimension of systems, are derived. A heuristic greedy algorithm of (quasi) optimal collapse is presented and validated statistically. The technique is applicable for solving sparse systems over arbitrary rings (fields) with sign. © 2016 Elsevier Inc.

Czaja L.,Vistula University | Czaja L.,University of Warsaw
Fundamenta Informaticae | Year: 2014

Petri net structures are used as communication model of network systems where message transfer channels are represented by the nets' edges, communicating agents - by nets' places and actions of communication - by nets' transitions, which here are called transmissions. The role of (structured) tokens play send/receive statements, that arrive at random, which makes the distribution of edges change dynamically. In this sense a net is self-modyfying: although the set of agents-places is fixed, the channels and communicating actions vary in the course of the net activity. Problems of deadlock and fairness is investigated.

Zaitsev D.A.,Vistula University
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

Finite classical Petri nets are non-Turing-complete. Two infinite Petri nets are constructed which simulate the linear cellular automaton Rule 110 via expanding traversals of the cell array. One net is obtained via direct simulation of the cellular automaton while the other net simulates a Turing machine, which simulates the cellular automaton. They use cell models of 21 and 14 nodes, respectively, and simulate the cellular automaton in polynomial time. Based on known results we conclude that these Petri nets are Turing-complete and run in polynomial time. We employ an induction proof technique that is applicable for the formal proof of Rule 110 ether and gliders properties further to the constructs presented by Matthew Cook. © Springer International Publishing Switzerland 2015.

Czaja L.,Vistula University | Czaja L.,University of Warsaw
CEUR Workshop Proceedings | Year: 2016

A new protocol using vectors of global timestamps for mutual exclusion in systems with Distributed Shared Memory (DSM) is described and some of its properties proved.

Placzek S.,Vistula University
CEUR Workshop Proceedings | Year: 2016

Two important issues have to be dealt with when implement-ing the hierarchical structure [1] of the learning algorithm of an Artificial Neural Network (ANN). The first one concerns the selection of the gen-eral coordination principle. Three different principles are described. They vary with regard to the degree of freedom for first-level tasks. The sec-ond issue concerns the coordinator structure or coordination algorithm. The ANN learning process can be examined as a two-level optimization problem. Importantly all problems and sub-problems are unstructured minimization tasks. The article concentrates on the issue of the coor-dinator structure. Using the interaction prediction principle as the most suitable principle for two-level ANN structures, different coordinator tar-get functions are defined. Using classification task examples, the main dynamic characteristics of the learning process quality are shown and analyzed.

Czaja L.,Vistula University | Czaja L.,University of Warsaw
CEUR Workshop Proceedings | Year: 2015

Two observations in the matter of pictorial as well as formal presentation of some consistency in distributed shared memory are made. The first concerns geometric transformation of line segments and points picturing read/write operations, the second - converting partial order of the operations into linear order of their initiations and terminations. This allows to reduce serialization of the read/write operations as a whole to permutations of their beginnings and ends. Some draft proposals are introduced.

Loading Vistula University collaborators
Loading Vistula University collaborators