Marcel Cremene | Technical University of Cluj-Napoca (original) (raw)

Papers by Marcel Cremene

Web service composition is the process of aggregating a set of existing web services in order to ... more Web service composition is the process of aggregating a set of existing web services in order to create new functionality. Current approaches to automatic web service composition are based, in general, on various Artificial Intelligence (AI) techniques, in particular search algorithms. Automated planning is one of the most frequently used. A limitation of these approaches is that they do not involve learning from previous attempts in order to improve the planning process. A new approach for automatic web service composition, based on Reinforcement Learning, is proposed. The method is suited for problems that do not define a particular goal to be reached but a reward that has to be maximized.

Interactive Evolutionary Computation (IEC) community aims at reducing user's fatigue during an op... more Interactive Evolutionary Computation (IEC) community aims at reducing user's fatigue during an optimization task involving subjective criteria: a set of graphic potential solutions are simultaneously shown to a user which task is to identify most interesting solutions to the problem he had to solve. Evolutionary operators are applied to user choices expecting to produce better solutions. As traditional IEC ask the user to give a mark to each solution or to explicitly choose bests solutions with a mouse, we propose a new framework that uses in real time gaze information to predict which parts of a screen is more significant for a user. We can therefore avoid the user to explicitly choose which solutions are interesting for him. In this paper, we mainly focus on automatically ordering solutions shown on a screen given a gaze path obtained by an eye-tracker. We applied several supervised learning methods (SVM, neural networks…) on two different experiments. We obtain a formula that predict with 85% user choices. We demonstrate that decisive criterion is time spent on one solution and we show the independency between this formula and the experiment.

Secure Mobile-Cloud is a framework proposed to secure the data transmitted between the components... more Secure Mobile-Cloud is a framework proposed to secure the data transmitted between the components of a mobile cloud application. In addition, the framework, takes into account the following aspects: 1) the users options regarding the security level required for private data and 2) the device energy consumption. The framework includes several distributed components. Some of these components are deployed on the mobile device and some of them in Cloud. This paper is focused on the implementation of the Secure Mobile-Cloud framework components on the mobile device. A proof of concept Android prototype is proposed.

Personalized ambient intelligent systems should meet changes in user’s needs, which evolve over t... more Personalized ambient intelligent systems should meet changes in user’s needs, which evolve over time. We propose *BAM – * Behaviour Adaptation Mechanism, a neural-network based control system that is trained, supervised by user’s (affective) feedback in real-time. The system deduces the preferred behaviour, based on the detection of affective state’s valence (negative, neutral and positive) from facial features analysis. The neural network is retrained periodically with the updated training set, obtained from the interpretation of the user’s reaction to the system’s decisions. The *BAM mechanism is implemented in the Affective-aware Smart Home (ASH), a multi-agent and ontological context-based system. We implemented a simple example consisting of a control system to position the blinds according to the inside and outside light level. We investigated how many training examples, rendered from user’s behaviour, are required in order to train the neural network so that it reaches an acc...

Situations where honest people interact with dishonest people are ubiquitous. Problems emerge spo... more Situations where honest people interact with dishonest people are ubiquitous. Problems emerge spontaneously and leaders must face the situations accordingly. In any civilized society honest people respect the laws while dishonest people do not. Decision makers need to take proper measures in order to avoid emergence of social problems as a consequence of dishonest behavior. Studies proved that in order to discourage social dishonest behavior, punishment probability is more important than punishment severity. Honest/dishonest dynamics are analyzed within the Social Honesty game. Transition intervals for punishment probability are indentified and analyzed. The following paper illustrates how the punishment probability influences the outcome of interactions between players, using asynchronous models of activation.

Today's trend in education is to in based tasks with practical activities. Stu encouraged to ... more Today's trend in education is to in based tasks with practical activities. Stu encouraged to work in groups to solve complex collaboration skills. But classical grouping s used in eLearning environments where th physical interaction between parties. We intr based grouping method that takes into ac typologies and the neuro-linguistic programm Typologies are determined, according to methodology, with the RHETI test. The NLP p by an eye-tracking system, based on eye m Case studies show that groups created using th increased communication among the mem practical results. This is due to the memb which facilitates better collaboration. eLearning; student grouping; grouping strat ntegrate computer- udents are often x tasks and develop strategies can't be here's little or no oduce a computer- ccount the student ming (NLP) profile. the Enneagram profile is evaluated movement patterns. his method show an mbers and better bers' compatibility,

Chaos, Solitons & Fractals, 2021

Abstract We consider a probabilistic-payoff social dilemma game and we analyze the impact of vari... more Abstract We consider a probabilistic-payoff social dilemma game and we analyze the impact of various social learning mechanisms - conformity, diversity generators, identity, on the dynamics of honest behavior in a virtual population. To capture the inherent variability and uncertainty of a complex social environment, the proposed model considers probabilistic payoffs as well as a variety of strategy updating rules and hybrid neighborhood topologies. The implicit underlying mechanisms are those of imitation and social influence. By simulating evolutionary spatial scenarios, we identify factors that favor contagion of honest behavior (i.e. social honesty) in the context of probabilistic payoff. Conformity and identity have a significant effect on the contagion of honest behavior. They slow down the dynamics, make it more stable, and help the formation and growth of clusters, favoring the emergence and spreading of honest behavior. However, their impact is topology dependent, nonlinear, exhibiting phase shifts. High conformity favors honest behavior in simple topologies, and does the opposite in complex topologies. Without a certain degree of diversity (use of alternative strategies) the system loses its adaptation capability. At the other end, strategy inconsistency leads to chaos, which favors the contagion of dishonest behavior. A structuring effect of leaders on chaotic behavior is also observed: their presence reduces the effects of random behavior, inducing alignment and reducing the necessity of a higher punishment probability. The key findings of this research can help predict trends and design public and organizational policies to nudge honest behavior, in the context of high complexity dynamics generated by the recent advances in information technology.

SpringerReference

The author's research activity, following the PhD defense in 2005, is presented from a unified in... more The author's research activity, following the PhD defense in 2005, is presented from a unified interdisciplinary perspective of complex adaptive systems. The Thesis subject lies at the crossroads of Software Engineering, Optimization Theory, Meta-heuristics, Evolutionary Computing and Game Theory. Service adaptation-a main topic of this Thesis-is a central concept in Software Engineering. Automatic complexity management represents an important challenge in fields such as Autonomic Computing and Self-Organizing Networks. Meta-heuristic based approaches offer efficient solutions for adaptation problems modeled as search/optimization problems. Multi-agent complex systems have interesting properties that provide powerful solutions to adaptation problems when a decentralized approach is adopted. Four main directions are investigated: a) designing a general model for the service adaptation problem based on a centralized control approach, b) proposing models and algorithms for complex decentralized distributed adaptive systems, c) analyzing different perspectives in decomposing/recomposing an adaptive system in/from several modules, and d) including psycho-social elements in the framework of technical systems. Complexity is considered here an orthogonal aspect related to all four directions stated before. The term complexity is used with twofold meaning: the first one is the computational complexity as it is used in Theoretical Computer Science; the second meaning of complexity is related to distributed decentralized complex systems composed of multiple autonomous entities. Complexity definitions, specific problems, properties, measures, and issues are discussed. A unique formulation that could resume my acknowledgments is: all the good ideas came to me by interacting with other people. Therefore, I want to thank all those whom I interacted with on scientific themes. The first place is reserved to all my co-authors. All these people are important yet there are some who had a special influence on my evolution: Professor Dan Dumitrescu offered me the great opportunity to collaborate with him and his research team. Evolutionary optimization and Game Theory tools came to me this way. Besides the numerous fruitful discussions, I also learned how important it is to stimulate and encourage people. My wife, Ligia, is also a great scientific collaborator. The most important aspect for which I should thank her is the ability to make unexpected connections between very different concepts from different fields and to find interesting analogies. She was at the origin of the interest in Cognitive Radio and socio-technical systems. Thanks are due to professor Costin Miron for the very useful interactions that continued after my Ph.D. studies. The most important thing learned from these interactions is the use critical thinking and a systematic intellectual approach. This is how I discovered the great importance of asking questions, looking from different perspectives, and managing time in small intervals. I also thank professor Michel Riveill for the collaboration after my PhD. An important number of results have been published in collaboration with his team. These interactions exposed me to high quality research. I received a very important support for the research as ATER and invited professor at

Telecommunication Systems, 2017

The paper proposes a general game theoretical model, called capacity demand game, for treating si... more The paper proposes a general game theoretical model, called capacity demand game, for treating simultaneous capacity requests in scarce-resource cognitive radio (CR) environments. The approach is that of non-cooperative games describing CR interactions in terms of radio resource access. Experiments reveal stable states (equilibria) that favour an equitable usage of radio resources to the benefit of all participants. Several equilibria are detected and discussed: Nash (NE), Pareto, joint Nash–Pareto, and Lorenz equilibrium.

International Journal of Computers Communications & Control, 2010

The widespread of Web services in the ubiquitous computing era and the impossibility to predict a... more The widespread of Web services in the ubiquitous computing era and the impossibility to predict a priori all possible user needs generates the necessity for on-demand service composition. Natural language is one of the the easiest ways for a user to express what he expects regarding a service. Two main problems need to be solved in order to create a composite service to satisfy the user: a)retrieval of relevant services and b) orchestration/composition of the selected services in order to fulfill the user request. We solve the first problem by using semantic concepts associated with the services and we define a conceptual distance to measure the similarity between the user request and a service configuration. Retrieved services are composed, based on aspect oriented templates called Aspects of Assembly. We have tested our application in an environment for pervasive computing called Ubiquarium, where our system composes a service according to the user request described by a sentence....

International Journal of Computers Communications & Control, 2011

Long-term adaptation solutions do not receive much attention in the design phase of a wireless sy... more Long-term adaptation solutions do not receive much attention in the design phase of a wireless system. A new approach is proposed, where the antenna takes an active role in characterising and learning the operation environment. The proposed solution is based on a signal fishing mechanism. Several software components, among which a genetic optimizer, implement the processing stages of autonomous design of the antenna array during operation.

Human centred services are increasingly common in the market of mobile devices. However, affectiv... more Human centred services are increasingly common in the market of mobile devices. However, affective aware services are still scarce. In turn, the recognition of secondary emotions in mobility conditions is critical to develop affective aware mobile applications. The emerging field of Affective Computing offers a few solutions to this problem. We propose a method to deduce user’s secondary emotions based on context and personal profile. In a realistic environment, we defined a set of emotions common to a museum visit. Then we developed a context aware museum guide mobile application. To deduce affective states, we first used a method based on the user profile solely. Enhancement of this method with machine learning substantially improved the recognition of affective states. Implications for future work are discussed

2013 11th RoEduNet International Conference, 2013

2013 RoEduNet International Conference 12th Edition: Networking in Education and Research, 2013

The mobility has become a valuable necessity for information technology users. Along with this pr... more The mobility has become a valuable necessity for information technology users. Along with this property, for a smart phone user, the following aspects are also important: the storage space, the energy consumption, the applications complexity and the efficiency. Comparing to standalone mobile applications, mobile cloud applications make use of resource sharing, have richer functionality and reduced costs. On the other hand, mobile cloud applications raise some problems related to data security, data privacy and trust. In this paper, we analyze the need for mobile cloud applications, we propose a scenario for an e-health application and we design a data traceability service.

Lecture Notes in Computer Science, 2013

Small-cell, open capacity sharing scenarios in Cognitive Radio (CR) environments are studied from... more Small-cell, open capacity sharing scenarios in Cognitive Radio (CR) environments are studied from a game theoretical (GT) perspective. Simultaneous capacity requests in small-cell scenarios are modelelled as strategic interactions between CRs and analysed as resource access games. CR capacity access competition is modelled based on discrete reformulations of the Bertrand GT model. Detected equilibria describe stable game situations. Numerical simulations identify situations where Nash equilibrium (NE) is both fair and Pareto efficient or where there are multiple NE solutions to choose from, indicating a flexible range for CR strategies. Adding to the analysis are the joint Nash-Pareto solutions (intermediate between Nash and Pareto) capturing heterogeneous behaviour of players. Stable and equitable states are detected even when players have different biases.

Proceedings of the 15th annual conference companion on Genetic and evolutionary computation, 2013

ABSTRACT When solving a multi-objective problem Pareto based evolutionary algorithms are the pref... more ABSTRACT When solving a multi-objective problem Pareto based evolutionary algorithms are the preferred choice. They are able to find a good approximation of the Pareto front and assure good diversity. But Pareto dominance scales badly with the number of objectives. Decomposition based algorithms represent a good choice for many-objective problems, their performance is not affected in such a severe way because they solve multiple one-objective problems. The preferred methods for scalarizing all objectives into one single objective are weighted sum and weighted Tchebycheff. With some modifications to the Tchebycheff approach some drawbacks, such as obtaining weak Pareto optimal solutions, can be avoided. We study the augmented, modified Tchebycheff and Lp decomposition techniques as an alternative. Numerical results on test problems indicate an in crease in performance over weighted sum and weighted Tchebycheff when applied to many-objective optimization problems.

Biomedical Engineering / 765: Telehealth / 766: Assistive Technologies, 2012

Studies in Computational Intelligence, 2011

International Journal of Computers Communications & Control, 2012

Service adaptation is one of the main research subjects in Ubiquitous Computing. Dynamic service ... more Service adaptation is one of the main research subjects in Ubiquitous Computing. Dynamic service adaptation, at runtime, is necessary for services that cannot be stopped (banking, airport, etc.). The classical approaches for dynamic adaptation require predicting all service and context states in order to specify service and context-specific adaptation policies. This prediction may lead to a combinatorial explosion. The aim of this research is to create a service and context-independent adaptation mechanism. Our proposal is based on a service-context model that is causally connected with the service and context, in a model@run.time paradigm. A closed-loop control principle is used for the adaptation mechanism. We introduce an equivalent for the error that is expressed by the notion of service-context distance. This distance represents a measure of how adequate is a service to its context. This distance is computed by some generic, reusable components. The adaptation algorithm that minimizes this distance is also service and context-independent.