Increasing societal demand for automation has led to considerable efforts to control large-scale complex systems, especially in the area of autonomous intelligent control methods. The control system of a large-scale complex system needs to satisfy four system level requirements: robustness, flexibility, reusability, and scalability. Corresponding to the four system level requirements, there arise four major challenges. First, it is difficult to get accurate and complete information. Second, the system may be physically highly distributed. Third, the system evolves very quickly. Fourth, emergent global behaviors of the system can be caused by small disturbances at the component level. The Multi-Agent Based Control (MABC) method as an implementation of distributed intelligent control has been the focus of research since the 1970s, in an effort to solve the above-mentioned problems in controlling large-scale complex systems. However, to the author's best knowledge, all MABC systems for large-scale complex systems with significant uncertainties are problem-specific and thus difficult to extend to other domains or larger systems. This situation is partly due to the control architecture of multiple agents being determined by agent to agent coupling and interaction mechanisms. Therefore, the research objective of this dissertation is to develop a comprehensive, generalized framework for the control system design of general large-scale complex systems with significant uncertainties, with the focus on distributed control architecture design and distributed inference engine design. A Hybrid Multi-Agent Based Control (HyMABC) architecture is proposed by combining hierarchical control architecture and module control architecture with logical replication rings. First, it decomposes a complex system hierarchically; second, it combines the components in the same level as a module, and then designs common interfaces for all of the components in the same module; third, replications are made for critical agents and are organized into logical rings. This architecture maintains clear guidelines for complexity decomposition and also increases the robustness of the whole system. Multiple Sectioned Dynamic Bayesian Networks (MSDBNs) as a distributed dynamic probabilistic inference engine, can be embedded into the control architecture to handle uncertainties of general large-scale complex systems. MSDBNs decomposes a large knowledge-based system into many agents. Each agent holds its partial perspective of a large problem domain by representing its knowledge as a Dynamic Bayesian Network (DBN). Each agent accesses local evidence from its corresponding local sensors and communicates with other agents through finite message passing. If the distributed agents can be organized into a tree structure, satisfying the running intersection property and d-sep set requirements, globally consistent inferences are achievable in a distributed way. By using different frequencies for local DBN agent belief updating and global system belief updating, it balances the communication cost with the global consistency of inferences. In this dissertation, a fully factorized Boyen-Koller (BK) approximation algorithm is used for local DBN agent belief updating, and the static Junction Forest Linkage Tree (JFLT) algorithm is used for global system belief updating. MSDBNs assume a static structure and a stable communication network for the whole system. However, for a real system, sub-Bayesian networks as nodes could be lost, and the communication network could be shut down due to partial damage in the system. Therefore, on-line and automatic MSDBNs structure formation is necessary for making robust state estimations and increasing survivability of the whole system. A Distributed Spanning Tree Optimization (DSTO) algorithm, a Distributed D-Sep Set Satisfaction (DDSSS) algorithm, and a Distributed Running Intersection Satisfaction (DRIS) algorithm are proposed in this dissertation. Combining these three distributed algorithms and a Distributed Belief Propagation (DBP) algorithm in MSDBNs makes state estimations robust to partial damage in the whole system. Combining the distributed control architecture design and the distributed inference engine design leads to a process of control system design for a general large-scale complex system. As applications of the proposed methodology, the control system design of a simplified ship chilled water system and a notional ship chilled water system have been demonstrated step by step. Simulation results not only show that the proposed methodology gives a clear guideline for control system design for general large-scale complex systems with dynamic and uncertain environment, but also indicate that the combination of MSDBNs and HyMABC can provide excellent performance for controlling general large-scale complex systems.

Venomous animals have evolved a vast array of peptide toxins for prey capture and defense. Nature has evolved the venoms into a huge library of active molecules with high selectivity and affinity, which could be explored as therapeutics or serve as a template for drug design. The individual components of venom i.e. toxins are used in ion channel and receptor studies, drug discovery, and formulation of insecticides. ‘Venominformatics is a systematic bioinformatics approach in which classified, consolidated and cleaned venom data are stored into repositories and integrated with advanced bioinformatics tools and computational biology for the analysis of structure and function of toxins.’ Conus peptides (conopeptides), the main components of Conus venom, represent a unique arsenal of neuropharmacologically active molecules that have been evolutionarily tailored to afford unprecedented and exquisite selectivity for a wide variety of ion-channel subtypes and neuronal receptors. Ziconotide (ω-conotoxin MVIIa from Conus magus (Magician's cone snail)), is proven as an intrathecally administered N-type calcium channel antagonist for the treatment of chronic pain (U.S. Food and Drug Administration. Center for Drug Evaluation and Research) attesting to the pharmaceutical importance of Conus peptides. From the point of view of protein sequence and structure analysis, conopeptides can serve as attractive systems for the studies in sequence comparison, pattern extraction, structure–function correlations, protein–protein interactions and evolutionary analysis. Despite their importance and extensive experimental investigations on them, they have been hardly explored through in silico methods. The present thesis is perhaps the first attempt at deploying a multi-pronged bioinformatics approaches for studies in the burgeoning field of conopeptides. In the process of sequence-structure-function studies of conopeptides, we have created several sequence patterns of different conopeptide families and these have been accepted for inclusion in international databases such as PROSITE, the first pattern database to have been developed (http://www.expasy.org/prosite) and INTERPRO (http://www.ebi.ac.uk/interpro). More importantly, we have carried out extensive literature survey on the peptides for which we have defined the patterns to create PROSITE compatible documentation files (PDOC6004, PDOC60025 and PDOC60027). We have also created a series of sequence patterns and associated documentation filesof pharmaceutically promising peptides from plants and venomous animals (including O-conotoxin and P-conotoxin superfamily members) with knottin scaffold. Knottins provide appealing scaffolds for protein engineering and drug design due to their small size, high structural stability, strong sequence tolerance and easy access to chemical synthesis. The sequence patterns and associated documentation files created by us should be useful in protein family classification and functional annotation. Even though patterns might be useful at the family level, they may not always be adequate at the superfamily level due to hypervariability of mature toxins. In order to overcome this problem, we have demonstrated the applicationos of multi-class support vector machines (MC-SVMs) for the successful in silico classification of the mature conotoxins into their superfamilies. TheI- and J-conotoxin-superfamily members were analyzed in greater detail. On the basis of in silico analysis, we have divided the 28 entries previously grouped as I-conotoxin superfamily in UniProtKB/Swiss-Prot (release 49.0) into I1 and I2 superfamilies inview of their having two different types of signal peptides and exhibiting distinct functions. A comparative study of the theoretically modeled structure of ViTx from Conus virgo, a typical member of I2-conotoxin superfamily, reveals the crucial role of C-terminal region of ViTx in blocking therapeutically important voltage-gated potassium channels. Putative complexes created by us of very recently characterized J-superfamily conotoxin p11-4a with Kv1.6 suggest that the peptide interacts with negatively charged extracellular loops and pore-mouth of the potassium channel and blocks the channel by covering the pore as a lid, akin to previously proposed blocking mechanism of kM-conotoxin RIIIK from Conus radiatus to Tsha1 potassium channel. This finding provides a pointer to experimental work to validate the observations made here. Based on differences in the number and distribution of the positively charged residues in other conopeptides from the J-superfamily, we hypothesize different selectivity profile against subtypes of the potassium channels for these conopeptides. Furthermore, the present thesis reports the application of order-statistic filters and hydrophobicity profiles for predicting the location of membrane-spanning helices. The Proposed method is in particular effective for the class of helical membrane proteins, namely the therapeutically important voltage-gated ion channels, which are natural targets of several conotoxins. Our suggested ab initio approach is comparatively better than other spatial filters, confirming to the efficacy of including the concept of order or ranking information for prediction of TM helicdes. Such approaches should be of value for improved prediction performance including in large-scale applications. In addition, anlaysis has been carried out of the role of context in the relationship between form and function for the true PDB hits of some nonCys-rich PROSITE patterns. We have found specific examples of true hits of some PROSITE patterns displaying structural plasticity by assuming significantly different local conformation, depending upon the context. The work was carried out as a part of the research interest in our group in studying structural and other features of protein sequence patterns. The Contributions of the candidate to venominormatics include, creation of protein sequence patterns and information highlighting the importance of the patterns as gleaned from the lteratures for family classification: profile HMM and MC-SVMs for conotoxin superfamily classification; in silico characterization of I1 and I2 conotoxin superfamilies; studies of interaction with Kv1 channels of typical members of I2 and 3 conotoxin superfamilies and development of improved methods for detecting membrane-spanning helices. Chapter I starts with a brief account of venominformatics; bioinformatics for venoms and toxins. Chapter 2 presents a regular expression based classification of Conus peptides. Chapter 3 revisits the 28 entries previously grouped as I-conotoxin superfamily in UniProt Swiss-Prot knowledgebase (release 49.0) having four disulfide bonds with Cys arrangement C-C-CC-CC-C-C and they inhibit or modify ion channels of nerve cells. Chapter 4 describes pseudo-amino acid composition and MC-SVMs approach for conotoxin superfamily classification. Chapter 5 describes in silico detection of binding mode with Kv1.6 channel of J-superfamily conotoxin p114a from bermivorouos cone snail, Conus planorbis. Chapter 6 presents a comparative sequence-structure-function analysis of naturally occurring Cys-rich peptides having the Knottin or inhibitor cystine knot(ICK) scaffold, from different plants and venomous animals based on information available in the knottin database(http://knottin.cbs.cnrs.fr/). Chapter 7 describes the application of order-statistic filters and hydrophobicity profiles for detecting membrane-spanning helices. Chapter 8 describes the role of context in the relationship between form and function for the true PDB hits of some non Cys-rich PROSITE patterns. Chapter 9 summaries the important findings of the present studies on naturally occurring bioactive Cys-rich peptides with emphasis on Conus peptides and their interactions with respective target such as voltage-gated ion channels.

Abstract Ethernet based VPLS (Virtual Private LAN Service) is a transparent, protocol independent, multipoint L2VPN (Layer 2 Virtual Private Network) mechanism to interconnect remote customer sites over IP (Internet Protocol) or MPLS (Multiprotocol Label Switching) based provider networks. VPLS networks are now becoming attractive in many Enterprise applications, such as DCI (data center interconnect), voice over IP (VoIP) and videoconferencing services due to their simple, protocol-independent and cost efficient operation. However, these new VPLS applications demand additional requirements, such as elevated security, enhanced scalability, optimum utilization of network resources and further reduction in operational costs. Hence, the motivation of this thesis is to develop secure and scalable VPLS architectures for future communication networks. First, a scalable secure flat-VPLS architecture is proposed based on a Host Identity Protocol (HIP). It contains a session key-based security mechanism and an efficient broadcast mechanism that increase the forwarding and security plane scalability of VPLS networks. Second, a secure hierarchical-VPLS architecture is proposed to achieve control plane scalability. A novel encrypted label-based secure frame forwarding mechanism is designed to transport L2 frames over a hierarchical VPLS network. Third, a novel Distributed Spanning Tree Protocol (DSTP) is designed to maintain a loop free Ethernet network over a VPLS network. With DSTP it is proposed to run a modified STP (Spanning Tree Protocol) instance in each remote segment of the VPLS network. In addition, two Redundancy Identification Mechanisms (RIMs) termed Customer Associated RIMs (CARIM) and Provider Associated RIMs (PARIM) are used to mitigate the impact of invisible loops in the provider network. Lastly, a novel SDN (Software Defined Networking) based VPLS (Soft-VPLS) architecture is designed to overcome tunnel management limitations in legacy secure VPLS architectures. Moreover, three new mechanisms are proposed to improve the performance of legacy tunnel management functions: 1) A dynamic tunnel establishment mechanism, 2) a tunnel resumption mechanism and 3) a fast transmission mechanism. The proposed architecture utilizes a centralized controller to command VPLS tunnel establishment based on real-time network behavior. Hence, the results of the thesis will help for more secure, scalable and efficient system design and development of VPLS networks. It will also help to optimize the utilization of network resources and further reduction in operational costs of future VPLS networks. / Tiivistelmä Ethernet-pohjainen VPLS (Virtual Private LAN Service) on läpinäkyvä, protokollasta riippumaton monipisteverkkomekanismi (Layer 2 Virtual Private Network, L2VPN), jolla yhdistetään asiakkaan etäkohteet IP (Internet Protocol)- tai MPLS (Multiprotocol Label Switching) -yhteyskäytäntöön pohjautuvien palveluntarjoajan verkkojen kautta. VPLS-verkoista on yksinkertaisen protokollasta riippumattoman ja kustannustehokkaan toimintatapansa ansiosta tullut kiinnostavia monien yrityssovellusten kannalta. Tällaisia sovelluksia ovat esimerkiksi DCI (Data Center Interconnect), VoIP (Voice over IP) ja videoneuvottelupalvelut. Uusilta VPLS-sovelluksilta vaaditaan kuitenkin uusia asioita, kuten parempaa tietoturvaa ja skaalautuvuutta, optimaalista verkkoresurssien hyödyntämistä ja käyttökustannusten pienentämistä entisestään. Tämän väitöskirjan tarkoituksena onkin kehittää turvallisia ja skaalautuvia VPLS-arkkitehtuureja tulevaisuuden tietoliikenneverkoille. Ensin väitöskirjassa esitellään skaalautuva ja turvallinen flat-VPLS-arkkitehtuuri, joka perustuu Host Identity Protocol (HIP) -protokollaan. Seuraavaksi käsitellään istuntoavaimiin perustuvaa tietoturvamekanismia ja tehokasta lähetysmekanismia, joka parantaa VPLS-verkkojen edelleenlähetyksen ja tietoturvatason skaalautuvuutta. Tämän jälkeen esitellään turvallinen, hierarkkinen VPLS-arkkitehtuuri, jolla saadaan aikaan ohjaustason skaalautuvuus. Väitöskirjassa kuvataan myös uusi salattu verkkotunnuksiin perustuva tietokehysten edelleenlähetysmekanismi, jolla L2-kehykset siirretään hierarkkisessa VPLS-verkossa. Lisäksi väitöskirjassa ehdotetaan uuden Distributed Spanning Tree Protocol (DSTP) -protokollan käyttämistä vapaan Ethernet-verkkosilmukan ylläpitämiseen VPLS-verkossa. DSTP:n avulla on mahdollista ajaa muokattu STP (Spanning Tree Protocol) -esiintymä jokaisessa VPLS-verkon etäsegmentissä. Väitöskirjassa esitetään myös kaksi Redundancy Identification Mechanism (RIM) -mekanismia, Customer Associated RIM (CARIM) ja Provider Associated RIM (PARIM), joilla pienennetään näkymättömien silmukoiden vaikutusta palveluntarjoajan verkossa. Viimeiseksi ehdotetaan uutta SDN (Software Defined Networking) -pohjaista VPLS-arkkitehtuuria (Soft-VPLS) vanhojen turvallisten VPLS-arkkitehtuurien tunnelinhallintaongelmien poistoon. Näiden lisäksi väitöskirjassa ehdotetaan kolmea uutta mekanismia, joilla voidaan parantaa vanhojen arkkitehtuurien tunnelinhallintatoimintoja: 1) dynaaminen tunnelinluontimekanismi, 2) tunnelin jatkomekanismi ja 3) nopea tiedonsiirtomekanismi. Ehdotetussa arkkitehtuurissa käytetään VPLS-tunnelin luomisen hallintaan keskitettyä ohjainta, joka perustuu reaaliaikaiseen verkon käyttäytymiseen. Tutkimuksen tulokset auttavat suunnittelemaan ja kehittämään turvallisempia, skaalautuvampia ja tehokkaampia VLPS järjestelmiä, sekä auttavat hyödyntämään tehokkaammin verkon resursseja ja madaltamaan verkon operatiivisia kustannuksia.

Un graphe arête-colorié Gc est un graphe dont les arêtes sont coloriées par un ensemble de couleurs données. Un sous-graphe de Gc est dit proprement colorié s'il ne contient pas d'arêtes adjacentes de même couleur. Un graphe ou multigraphe c-arête-colorié Gc, est dit k-lié (respectivement k-arête-lié) si et seulement si quelque soient 2k sommets distincts de V(Gc), notés, x1 y1 , x2 y2 , ..., xk yk , il existe k chaînes élémentaires sommet-disjointes (respectivement arête-disjointes) proprement arête-coloriées, reliant x1 à y1 , x2 à y2 , ... , xk à yk .Un arbre couvrant propre d'un graphe Gc est un sous-graphe de Gc qui est un arbre couvrant proprement colorié.Un arbre couvrant faiblement colorié est une arborescence telle qu'il existe une chaîne proprement coloriée entre la racine et chaque sommet du graphe.Dans la première partie de cette thèse, nous donnons des conditions suffisantes pour qu'un graphe arête-colorié soit k-lié. C'est un problème classique en théorie des graphes, avec des applications multiples. Ainsi, nous avons établi entre autres les résultats suivants.A) Tout multigraphe 2-arête-colorié d'ordre n ≥ 242k tel que dc(Gc) ≥ n/2+k –1, est k-lié. B) Tout multigraphe c-arête-colorié d'ordre n ≥ 2k et de taille m≥ cn(n–1)/2 – c(n–2k +1)+1 est k-lié.C) Tout multigraphe c-arête-colorié d'ordre n ≥ 2k tel que dc(x) ≥ n/2 pour tout sommet x, est k-arête-lié.D) Tout multigraphe 2-arête-colorié d'ordre n ≥ 2k ≥ 10 et de taille m ≥ n2 –5n + 11 tel que dc(x) ≥ 1 pour tout sommet x, est k-arête-lié.Dans la seconde partie de cette thèse, deux autres problèmes classiques en théorie des graphes sont traités dans la version arête-coloriée. Il s'agit des arbres couvrants et des chaînes hamiltoniennes. Nous donnons ci-dessous quelques résultats.E) Tout graphe simple c-arête-colorié k-connexe d'ordre n ≥ C²k+1 + k + 2 avec c ≥ C²n–k–1 + k +1, a un arbre couvrant propre.F) Tout graphe Gc connexe c-arête-colorié de degré rainbow rd(Gc)=k et d'ordre n ≥ C²k+1 + k + 2 avec c ≥ C²n–k–1 + k +1, possède un arbre couvrant propre.G) Tout graphe simple c-arête-colorié k-connexe d'ordre n ≥ ((k + j)2 + 3(k + j) – 2)/2 avec c ≥ ((n – k – j)(n – k – j – 1))/2 + 2 , où j(j –1)=k , possède un arbre couvrant faiblement colorié.H) Tout multigraphe Gc d'ordre n ≥ 14 et de taille m ≥ (n – 3)(n – 4) + 3n – 2 tel que rd(Gc) = 2, possède une chaîne hamiltonienne propre. I) Tout multigraphe c-arête-colorié d'ordre n ≠ 5, 7 et de taille m ≥ n2 – 3n + 4, possède une chaîne hamiltonienne propre.La plupart des résultats exposés, sont les meilleurs possibles relativement aux propriétés sur les conditions suffisantes. / A c-edge-colored graph Gc is a graph whose edges are colored by a given set of colors. A subgraph of Gc is proper if no two adjacent edges have the same color.A c-edge-colored graph or multigraph Gc is k-linked (respectively k-edge-linked) if for any 2k distinct vertices, say x1 y1 , x2 y2 , ..., xk yk , there exist k vertex-disjoint (respectively edge-disjoint) proper paths joining x1 to y1 , x2 to y2 , ... , xk to yk .A proper spanning tree of a graph Gc is a spanning tree such that any two adjacent edges differ in colors.A weak spanning tree is a spanning rooted tree such that there exists a proper path between the root and every vertex of the graph.In the first part of this thesis, we provide conditions which are sufficient for an edge-colored graph to be k-linked. It is a classic problem in graph theory , with many applications. So, we established among others the following results.A) Every 2-edge-colored multigraph of order n ≥ 242k such that dc(Gc) ≥ n/2+k –1, is k-linked.B) Every c-edge-colored multigraph of order n ≥ 2k and size m≥ cn(n–1)/2 – c(n–2k +1)+1 is k-linked.C) Every c-edge-colored multigraph of order n ≥ 2k is k-edge-linked if for each vertex x, dc(x) ≥ n/2.D) Every 2-edge-colored multigraph of order n ≥ 2k ≥ 10 and size m ≥ n2 – 5n + 11 is k-edge-linked if for each vertex x, dc(x) ≥ 1.In the second part of this thesis, two other classic problems in graph theory are treated in edge-colored version: spanning trees and hamiltonian paths. We give below some results.E) Every c-edge-colored simple k-connected graph of order n ≥ C²k+1 + k + 2 with c ≥ C²n–k–1 + k +1, has a proper spanning tree.F) Every c-edge-colored connected graph Gc of rainbow degree rd(Gc)=k and order n ≥ C²k+1 + k + 2 with c ≥ C²n–k–1 + k +1, has a proper spanning tree. G) Every c-edge-colored simple k-connected graph of order n ≥ ((k + j)2 + 3(k + j) – 2)/2 and c ≥ ((n – k – j)(n – k – j – 1))/2 + 2 , with j(j –1)=k , has a weak spanning tree.H) Every c-edge-colored multigraph Gc of order n ≥ 14 and size m ≥ (n – 3)(n – 4) + 3n – 2 such that rd(Gc) = 2, has a proper hamiltonian path.I) Every c-edge-colored multigraph of order n ≠ 5, 7 and size m ≥ n2 – 3n + 4, has a proper hamiltonian path.Most of the given results are the best possible with regard to the properties on the sufficient conditions.

Dödligt skolvåld har sedan länge varit ett uppmärksammat problem i bland annat USA. Forskningen om dödligt skolvåld tenderar ofta att studera amerikanska exempel där syftet är att identifiera riskindivider genom att undersöka personlighetsdrag och bakgrundsfaktorer. Dödligt skolvåld i en skandinaviskt kontext saknas det dock kunskap om. Syftet med den här studien är att undersöka utmaningar och framgångsfaktorer med samverkan vid hanteringen före, under och efter händelser med dödligt våld i den svenska skolmiljön. Studien genomfördes med hjälp av kvalitativ metod där det empiriska materialet samlades in genom semistrukturerade intervjuer. Med utgångspunkt i ledning och samverkan som teoretiskt ramverk och med boundary spanning som centralt teoretiskt begrepp, analyserades insamlade data med tematisk analys. Studien resulterade i två huvudsakliga teman, utmaningar för samverkan och framgångsfaktorer för samverkan. Studien visade att avsaknad av systematik och bristande kommunikation utgör de främsta utmaningarna, samtidigt som god person- och förmågekännedom utgör de främsta framgångsfaktorerna vid samverkan. Studiens resultat bekräftar delvis tidigare forskning men lyfter också fram nya fynd. Till exempel att lagstadgade sekretessbestämmelser utgör hinder för effektiv samverkan mellan skolan, polisen och socialtjänsten. / Deadly school violence has long been acknowledged as a pressing issue, particularly within the United States. Existing research primarily concentrates on American instances, with a focus on identifying high-risk individuals through the examination of personality traits and background factors. However, there exists a notable gap in understanding deadly violence within a Scandinavian context. This study aims to explore the challenges and success factors associated with collaborative efforts in managing incidents of deadly violence within the Swedish school environment, spanning the periods before, during, and after such occurrences. Employing a qualitative approach, empirical data was gathered through semi-structured interviews. Drawing upon the theoretical framework of management and collaboration, and boundary spanning as a central theoretical concept, thematic analysis was employed to analyze the collected data. The study highlights two key themes: challenges for collaboration and success factors for collaboration. Findings reveal that the lack of systematic approaches and inadequate communication represent primary challenges, while good understanding of individuals and their capabilities are the primary success factors in collaboration. The study's findings partly confirm previous research but also highlight new discoveries. For instance, statutory confidentiality regulations pose obstacles to effective collaboration between schools, the police, and social services.

This qualitative research addressed systems psychodynamic consultation to boundary management. The systemic, dynamic and chaotic aspects of organisational life formed the backdrop against which the research was conducted. The general objective of the research was to describe a relevant consulting model for organisational consulting psychologists related to boundary management. Literature was reviewed in order to describe organisational consulting and organisational boundaries from the systems psychodynamic perspective. Key principles for boundary management consulting were also described. The objectives of the empirical study were to apply psychodynamic consulting to boundary management and to describe the process. A further objective was to produce research hypotheses about boundary management from both an organisational and a consulting perspective. A case study design was followed. Descriptive data was gathered by means of a participative observer. The data was analysed by means of systems psychodynamic discourse analysis. Ten working hypothesis were produced. These hypotheses culminated into two research hypotheses, describing the primary task of boundary management and boundary management consulting. The first research hypothesis was that the primary task of boundary management is to hold the polarities of integration and differentiation, not allowing the system to become fragmented or overly integrated. The second research hypothesis was that the primary task of the consultant in boundary management consulting is to help the organisation’s managing its own boundaries. This is carried out through taking up the role of organisational consultant, performing the consulting tasks and by applying a consulting process. The researcher concluded that boundary management is an activity of the whole organisation. Boundary management consulting facilitates or supports this organisational activity.

Esta dissertação trata do problema da k-árvore de custo mínimo (kMST): dados um grafo conexo G, um custo não-negativo c_e para cada aresta e e um número inteiro positivo k, encontrar uma árvore com k vértices que tenha custo mínimo. O kMST é um problema NP-difícil e portanto não se conhece um algoritmo polinomial para resolvê-lo. Nesta dissertação discutimos alguns casos em que é possível resolver o problema em tempo polinomial. Também são estudados algoritmos de aproximação para o kMST. Entre os algoritmos de aproximação estudados, apresentamos a 2-aproximação desenvolvida por Naveen Garg, que atualmente é o algoritmo com melhor fator de aproximação. / This dissertation studies the minimum cost k-tree problem (kMST): given a connected graph G, a nonnegative cost function c_e for each edge e and a positive integer k, find a minimum cost tree with k vertices. The kMST is an NP-hard problem, which implies that it is not known a polynomial algorithm to solve it. In this dissertation we discuss some cases that can be solved in polynomial time. We also study approximation algorithms for the kMST. Among the approximation algorithms we present the 2-approximation developed by Naveen Garg, which is currently the algorithm with the best approximation factor.

Cette thèse s’inscrit dans le cadre de l’évolution des Systèmes d’Informations Géographiques (SIG) et de leur capacité à répondre à des problématiques environnementales qui s’expriment de manière globale et transversale. Dans un contexte ou l’information géographique est en plein essor et où la quantité de données disponible ne cesse de croitre, le besoin en terme d’outil d’aide a la décision n’a jamais été aussi fort. Cette étude s’attache tout particulièrement au cadre de la résolution de problématiques liées à l’eau et l’environnement lorsque les données deviennent trop volumineuses pour être traitées par des moyens de calculs séquentiels classiques. Nous proposons une plateforme de calculs répartis sur une grappe d’ordinateurs qui parallélise le calcul de la délimitation des bassins versants des grands fleuves et la détermination des flots d’accumulation. A cette fin nous introduisons une technique de calcul parallèle d’une forêt d’arbres couvrants minimums représentant le parcours de l’eau de chaque point du Modèle Numérique de Terrain (MNT) vers la mer. Cette technique débute par une délimitation des cuvettes (ensemble de points allant vers le même minimum local) contenues dans le MNT. Ensuite une hiérarchie de déversement des cuvettes les unes dans les autres est construite jusqu'à obtenir les bassins versants des fleuves. L’étude se poursuit par la description d’un algorithme parallèle pour le calcul très couteux des flots d’accumulation en chaque point du MNT. Enfin cette thèse présente une version ≪out-of-core≫ de nos algorithmes parallèles afin d’étendre la portée de nos travaux a des grappes de dimensions modestes qui ne peuvent pas charger en mémoire la totalité du MNT traite. / My thesis is part of the development of Geographic Information Systems (GIS) and their ability to respond to environmental challenges that are expressed in a global and transversal way. We consider a context in which geographical information is growing, in addition the amount of data available continues to grow. Therefore, the need a tool for decision support has never been stronger. This study aim to solve problems related to water and the environment when the data become too large for sequential computing. The main objective of the thesis proposes a platform for distributed computing on a cluster of computers that parallelizes the watershed computing of major rivers and the determination of the flow accumulation. The idea is based on the construction of a minimal spanning tree, via a hierarchy of graphs, modeling the water route on the DEM toward the ocean. The technique begins from computing catchment basins that are set of pixels for which a drop of water will end the same local minimum. After that, a hierarchy of basins is computed in order to give the catchment basins of the rivers in the DEM. The study continues with a description of a parallel algorithm for computing the global flow accumulation for automatic drainage network extraction in large digital elevation models. Finally, the thesis presents a version ≪out-of-core≫ of our parallel algorithms to extend the scope of our work in clusters of size small that cannot load into memory the entire treated DEM.

Esta dissertação trata do problema da k-árvore de custo mínimo (kMST): dados um grafo conexo G, um custo não-negativo c_e para cada aresta e e um número inteiro positivo k, encontrar uma árvore com k vértices que tenha custo mínimo. O kMST é um problema NP-difícil e portanto não se conhece um algoritmo polinomial para resolvê-lo. Nesta dissertação discutimos alguns casos em que é possível resolver o problema em tempo polinomial. Também são estudados algoritmos de aproximação para o kMST. Entre os algoritmos de aproximação estudados, apresentamos a 2-aproximação desenvolvida por Naveen Garg, que atualmente é o algoritmo com melhor fator de aproximação. / This dissertation studies the minimum cost k-tree problem (kMST): given a connected graph G, a nonnegative cost function c_e for each edge e and a positive integer k, find a minimum cost tree with k vertices. The kMST is an NP-hard problem, which implies that it is not known a polynomial algorithm to solve it. In this dissertation we discuss some cases that can be solved in polynomial time. We also study approximation algorithms for the kMST. Among the approximation algorithms we present the 2-approximation developed by Naveen Garg, which is currently the algorithm with the best approximation factor.

Spinal cord injury (SCI) is a devastating disease which poses health problems in human and veterinary medicine. SCI causes neurological disability, with loss of motor, sensory and autonomic function. This study investigated the efficacy of local treatment with IKVAV-peptide on spinal cord regeneration following compression injury at T12 vertebra in Balb-c mice. IKVAV-peptide is a membrane spanning peptide known to have a long half-life and the peptide motif IKVAV. Thirty Blab-c female mice were used. Hemilaminectomy was performed at T12 and spinal cords were compressed using extradural application of a 24 g modified aneurysm clip for 1 min in the treatment groups. After 24 hours mice were treated with one of 4 different treatments including isoleucine-lysine-valine-alanine-valaine(IKVAV), IKVAVpeptide, peptide and mannitol (vehicle). Functional improvement was assessed every day using Basso, Beattie, Bresnahan (BBB) Locomotor Rating Scale. 28 days later, the mice were euthanized, and spinal cord segments were studied histologically. Statistical analysis, one-way and two-way analysis of variance (ANOVA) and linear regression model were used to measure some parameters and describe the outcome after SCI. Over a 4weeks period, IKVAV-peptide group demonstrated statistical and histological evidence of cellular reconstruction and behavioral improvement. The BBB score in the IKVAV-peptide group increased by 5.4 (25%) points, the IKVAV and peptide groups by approximately 1 point (5%) and the mannitol group by 4 points (19%). The number of protoplasmic astrocytes in the IKVAV-peptide group was significantly increased compared to IKVAV, mannitol and normal groups but not with the peptide group (p<0.001). Neuron and muscle bundle size were also increased significantly (p<0.05 and p<0.007, resp.) in the IKVAV-peptide group compared to other treatment groups. The treated control groups showed cellular and gross damages including neuron inactivation and muscle atrophy, gliosis and inability of movement. / Graduation date: 2013