Volume 19, Number 12, 2016
Dan DASCĂLU
Linqiang PAN, Gheorghe PĂUN, Tingfang WU, Zhiqiang ZHANG Abstract. Two research directions on numerical P systems and two onspiking neural P systems which were recently explored are shortly discussed. They deal with very natural versions of these classes of P systems which were not investigated before. Besides definitions, one recalls a few of the results obtained about these new classes of P systems and one points out open problems and directions for further research.
Alberto LEPORATI, Luca MANZONI, Giancarlo MAURI, Antonio E. PORRECA, Claudio ZANDRON Abstract. It has been recently proved that polynomialtime tissue P systems with cell division are only able to solve decision problems in the complexity class P when their cell structure is embedded into the Euclidean space , for . In this paper we show that if the space has an appropriate shape and is polynomialtime navigable (but not embeddable in ), then it is possible to even solve PSPACEcomplete problems. This means that the computational power of tissue P systems can be varied from P to PSPACE by just operating on the properties of the space in which they are located.
Dan CRISTEA, Andreea MACOVEI Abstract. We present in this paper a proposal for representing temporality phenomena as expressed by specific free writing techniques, such as temporal discontinuities, transitions from one story to another, backs and change of perspective, phenomena naturally understood by a casual human reader of the text. The long time intention is to develop a technique capable to reproduce on a machine the human reader performances with a sufficiently high accuracy. The diagrams we draw, called time yards, and the associated XML notations include: time tracks, themselves being made out of time segments, the atoms of time segments being the events. The paper makes a preliminary evaluation of the objectivity of recognizing these concepts as catched by a group of annotators.
Gheorghe TECUCI, Mihai BOICU, Dorin MARCU, David A. SCHUM Abstract. Evidencebased reasoning which is at the core of many problem solving and decision making tasks in a wide variety of domains, requires a complex combination of imaginative reasoning, critical reasoning, and expert knowledge. This paper presents research performed in the Learning Agents Center of George Mason University on developing a computational theory of evidencebased reasoning and on its integration within the Disciple learning agents theory, methodology, and tools for the development of cognitive assistants that are able to learn complex problem solving expertise directly from human experts, assist experts and nonexperts with problem solving and decision making, and teach the acquired expertise to students. The paper introduces evidencebased reasoning concepts that need to be understood in order to use such a cognitive assistant, illustrating them in the context of a specific system developed for intelligence analysis, named Cogent. It also illustrates a sample session with Cogent and summarizes the assistance it provides to the user. The paper concludes with the envisioned application of these results to other evidencebased reasoning domains.
Claudiu BRANDAS, Ciprian PANZARU, Florin Gheorghe FILIP Abstract. The paper reviews recent advances in the domain of computer supported data driven decisionmaking and proposes a conceptual framework for the construction and implementation of a class of practical decision support systems named Labour Market Decision Support Systems (LMDSS) to be applied to labour market analysis. LMDSS are meant to support the analysis the information and knowledge about dynamics of job vacancies and skills, concentration on territorial units and economic activities, most requested skills, education and training programmes, unemployment and labour force statistics, wage information and projections. The experimental results are presented in the context af academia labour market.
Dan TUFIȘ Abstract. Over the last twenty years or so, the approaches to partofspeech tagging based on machine learning techniques have been developed or ported to provide highaccuracy morpholexical annotation for an increasing number of languages. Given the large number of morpholexical descriptors for a morphologically complex language, one has to consider ways to avoid the data sparseness threat in standard statistical tagging, yet to ensure that the full lexicon information is available for each wordform in the output. The paper overviews some of the major approaches to partofspeech tagging and touches upon the tagset design, which is of crucial importance for the accuracy of the process. Read the pdf
Ioan DUMITRACHE, Dragoș C. POPESCU Abstract. In this paper is presented a new hybrid methodology using a Generalized Neuron with a Genetic Algorithm. After a short presentation of the capabilities of GN is shown a possibility to optimize the parameters by using a GA. As a case study is proven the efficiency of this new GNGA architecture in optimization of the estimation for battery State of Charge. Are presented some simulation results in comparison with the conventional Artificial Neural Networks.
HoriaNicolai TEODORESCU, Stefan Wolfgang PICKL Abstract. The paper addresses the concept of resilience as a two joint processes and focuses especially on the recovery phase. The treatment uses the frame of the resilience index previously introduced. The main contributions are the results on resilience index properties and optimization and the application of graph measures in the calculation of the recovery phase for resilience definition. The resilience index computation is exemplified for information systems.
Octavian PASTRAVANU, MihaelaHanako MATCOVSCHI, Mihail VOICU Abstract. Consider an arbitrary switching positive system whose (discrete or continuoustime) dynamics is defined by the matrix set A. The class of copositive Lyapunov functions called “polyhedral” is characterized by necessary and sufficient conditions derived from the solvability of some inequalities built with the matrices in A, and from the properties of the row representatives of A. Duality is proven between the “polyhedral” class and the already known class of “linear” copositive Lyapunov functions (characterized by equivalent conditions derived from the properties of the column representatives of A). A numerical example illustrates the utilization of two types of approaches, as well as the connection among them.
Octavian PASTRAVANU, MihaelaHanako MATCOVSCHI, Mihail VOICU Abstract. The paper studies the set invariance properties exhibited by the linear dynamics of interval matrix systems (abbreviated IMSs). The considered sets have nonsymmetrical shapes defined by arbitrary Hölder pnorms and are exponentially contractive (with exponential timedecrease). For such sets we provide two types of results that allow invariance tests: (i) sufficient conditions for , (ii) necessary and sufficient conditions for . Our approach includes, in generalized forms, results previously reported by other works referring to either nonsymmetrical sets invariant with respect to singlemodel systems, or symmetrical sets invariant with respect to IMSs. Our presentation focuses on continuoustime IMSs, whereas the discretetime case is briefly addressed by the key elements of a basic analogy. An example already used in the IMS literature illustrates the applicability of our theoretical developments.
Gheorghe M. ȘTEFAN1, Călin BÎRA, Radu HOBINCU, Mihaela MALIȚA Abstract. The emergence of CPU/FPGA hybrid processors promotes the FPGA accelerators as circuits which perform the critical functions associated to an application. As circuits, they are designed and optimized by hardware designers or, with a debatable efficiency, by specialized compilers. We propose, as an intermediate solution, to use a predesigned parameterized family of programmable generic structures, ready to be synthesized in FPGA technology as highly parallel computing engines able to perform efficiently data intense computations. The solution loses, in the worst cases, tens of percent of the processing power, compared with optimized hardware design, but gains a lot in flexibility allowing the number of skilled users to increase with few orders of magnitude. The proposed solution supposes to select, from a family of predesigned structures, an appropriate RTL module, to implement it in FPGA, and to program the resulting parallel engine in a highlevel programming language by a software designer. The theoretical foundation for the generic family of parallel engines is grounded on the synergy between Stephen Kleene’s computational model of Partial Recursive Functions and the Functional Programming Systems proposed by John Backus.
Andrei A. MÜLLER, Esther SANABRIACODESAL, Alin MOLDOVEANU, Victor ASAVEI, Dan DASCĂLU Abstract. The paper is describing and presenting the recent advances in the 3D Smith chart representations and applications and then it proposes a new conceptual model for the extended 2D Smith chart based on hyperbolic geometry by mapping the generalized Smith chart in the unit disc using the stereographic projection from a hyperboloid (Poincare disc model).
Mihai MIHAILA Abstract. Fluctuations with 1/f spectrum stand for the fundamental source of decoherence in the nanostructures used to process information in a quantum computer. Consequently, intrinsic 1/f noise of these nanodevices is acting as a roadblock in the field of quantum information and computing. This is another, very stringent reason why finding where this noise comes from became of great practical importance. However, understanding the microscopic mechanism of this ubiquitous phenomenon in solid remained one of the most difficult endeavors since its discovery more than 90 year ago. In this work, five different 1/f noiserelated issues are questioned and analyzed from a phenomenological perspective. The aim is to find possible, sofar hidden bridges between them, which might be helpful in grasping the microscopic origin of 1/f noise in solid. For instance, in deducing his celebrated fluctuationdissipation theorem, Nyquist assumed the validity of the energy echipartition in an ideal transmission line. We show that for a real system, it is necessary to clarify how the thermal energy is transferred between the eigenmodes. This can happen only if the modes are coupled by microscopic nonlinearities, whose presence confers a dynamic character to the equilibrium state. Therefore, nonlinearityinduced energy exchange between the modes is a condition necessary to have equilibrium fluctuations at the resistor terminals. Thereafter, the empirical procedure Hooge used to get the 1/f noise formula is analyzed. Two unexpected ideas were found behind it. The first one is that 1/f noise cannot exist in the absence of thermal noise, which would suggest an equilibrium microscopic mechanism for 1/f noise. The second one relates to the elimination of the kBT factor from the thermal noise spectrum. Since kBT factor is the echipartition fingerprint in thermal noise, its elimination appears as equivalent to the assumption that violation of equipartition is necessary to have 1/f noise. Next, it is shown that deviations of the frequency exponent from 1 must mirror those of the voltage exponent from 2, which means nonvalidity of the Ohm law. Therefore, any frequency exponent different from 1 should be considered as the possible macroscopic signature of a microscopic nonlinearity. Examples of deviations from linearity of the 1/f noise for semiconductors, metals and carbonic materials are given. The experiment done by Voss and Clarke to demonstrate that 1/f noise exists in thermal noise is very briefly described. Its explanation relies on the presence of microscopic nonlinearities. It resulted that the microscopic nonlinearity is a sine qua non condition for the existence of both thermal and 1/f noise. The FermiPastaUlam paradigmatic experiment is briefly presented in the end. This experiment demonstrating that equipartition law is violated can be of significance to unveil the mechanism of 1/f noise in solid.
Victor COJOCARU, Anatol SIDORENKO, Daniel VRABII Abstract. Strokes are an acute neurological disorder resulting from blockage of blood supply to an area of the brain or cerebral hemorrhage. Strokes are a clinical syndrome produced by brain tissue damage following ischemic events. One effective method in combating the consequences of trauma is therapeutic hypothermia. Currently, there is an important direction in modern biotechnology and physiology associated with experimental and theoretical study of heat transfer processes in the body and its individual organs and tissues. In this brief study, we simulate the heat transport processes while performing therapeutic hypothermia by using the Heat Transfer Module of COMSOL MULTYPHYSICS™ software in view of improving the design of the device we built for hypothermia for therapeutic treatment.
