Dissertations / Theses on the topic 'Local Learning and Employment Networks'

In common with many Western nations, Australian governments, both state and federal, have increasingly embraced network-based approaches in responding to the effects of globalisation. Since 2001, thirty one Local Learning and Employment Networks (LLEN) have been established across all areas of Victoria, Australia in line with recommendations of a Ministerial Review into Post Compulsory Education and Training Pathways. That review reported that, in the globalised context, youth in transition from schooling to independence faced persistent and severe difficulties unknown to previous generations; it also found problems were frequently concentrated in particular groups and regions. LLEN bring together the expertise and experience of local education providers, industry, community organisations, individuals and government organisations. As a result of their local decisions, collaboration and community building efforts it is intended that opportunities for young people will be enhanced. My research was conducted within an Australian Research Council Linkage Project awarded to Deakin University Faculty of Education in partnership with the Smart Geelong Region LLEN (SGR LLEN). The Linkage Project included two separate research components one of which forms my thesis: a case study of SGR LLEN. My data was generated through participant observation in SGR LLEN throughout 2004 and 2005 and through interviews, reflective writing and archival review. In undertaking my analysis and presenting my thesis I have chosen to weave a series of panels whose orientation is poststructural. This approach was based in my acceptance that all knowledge is partial and fragmentary and, accordingly, researchers need to find ways that highlight the intersections in and indeterminacy of their empirical data. The LLEN is -by its nature as a network -more than the contractual entity that gains funding from government, acts as the administrative core and occupies the LLEN office. As such I have woven firstly the formation and operational structure of the bounded entity that is SGR LLEN before weaving a series of six images that portray the unbounded LLEN as an instance-in-action. The thesis draws its theoretical inspiration from the work of Deleuze and Guattari (1987). Despite increased use of notions of networks, local decision-making and community building by governments there had been little empirical research that explored stakeholder understandings of networks and their role in community building as well as a lack of theorisation of how networks actually ‘work.’ My research addresses this lack and suggests an instituted network can function as a learning community capable of fostering systemic change in the post compulsory education training and employment sector and thereby contributing to better opportunities for young people. However the full potential of the policy is undermined by the reluctance of governments to follow through on the implications of their policies and, in particular, to confront the limiting effects of performativity at all levels.

This study investigated the appropriate methods for incorporating collaborative activities supported by wireless local area networks in the middle school classroom environment. The project addressed the problem that although current educational practice indicates the value of incorporating collaborative learning in the middle school classroom, the physical setting, lack of training, and a narrow scope of vision tend to block that incorporation in many schools. The goal of this project was to develop and evaluate a process model that can be utilized to facilitate the collaborative learning process through the incorporation of wireless LAN's in a middle school environment. Criteria that are necessary for a successful product were established and matrices based on the established criteria were developed. An eight-member Formative Committee was assembled to establish a criteria set for the process model. The committee consists of seven educators from the Emanuel County School System, plus one representative from the system-wide technological support staff. The Formative Committee examined current practices and methodologies associated with collaborative learning in the classroom. This committee also investigated methods in which collaborative activities can be enhanced through the utilization of wireless networks. An Expert Committee was responsible for validating the criteria established by the Formative Committee. This four member Expert Panel was selected from teacher educators in local colleges and universities based on teaching experience, expertise in collaborative learning, expertise incorporating computers in the classroom, and expertise developing educational support materials. The validation was accomplished utilizing surveys and formative feedback based on criteria recommended by the Formative Committee. Current collaborative activities were examined and guidelines were developed that meet the requirements of the validated criteria. A pilot study was conducted in order to aid in the developmental process. An eighth grade music appreciation course was utilized in the pilot study. This class utilized computers connected via a wireless network and the process model that was developed based on the established criteria, to support the collaborative activities. A Development Committee examined the results from the pilot study. This committee was composed of three members from the Formative Committee and three members from the Expert Panel. Data gathered from the Development Committee and the pilot study was incorporated into the process model. After recommendations from the Development Committee were examined and integrated into the product, a Summative Committee was developed in order to evaluate the process model. This committee presented recommendations for future modifications as well as future areas of research.

This dissertation is an attempt of expanding the domain of game theory into the sphere of evolving, potentially non-equilibrium systems. We especially focus our attention on studying the effects of local interactions, using automata networks as a modelling tool. The Chapters 2 and 3 of this dissertation concentrate on applications of the local nature of interactions and rely on automata networks as an investigating and modelling tool for game theory. Chapter 2 is devoted to cooperation and to a smaller extent to the endogenous formation of links between the agents. Chapter 3 is investigating the deterministic and stochastic best response play when interactions are local.
Ph. D.

This thesis focuses on the practice of local learning centres in Sweden. The aim is to describe and to establish an understanding of relations and the actor-networks that surround the practice. The thesis is based on four different studies. The first two studies focus on the organisation of the local learning centres and scrutinises the relation between the local learning centres, municipal administration and business in the region. What types of relations exist and what strategies are used by the local learning centres to enrol actors into the actor-network? The two latter studies investigate videoconference in the context of local learning centres. Videoconference is described as a socio-technical environment. The studies focus on interaction and communications asking what kind of relations are created between environment, material design and people? The theoretical framework consists of actor-network theory and the notions of space and spatial relations. The methods used in the four studies are interviews and observations. Informants that have contributed to the study are headmasters, teachers and students as well as project managers, politicians and businessmen. The results of the studies show how the local learning centres fails as brokers on an educational market. The reasons are many. The main results of the studies stress a number of obstacles of involvement and shed light on different strategies of enrolment. The results connected to videoconference show how material design and technology impacts the practice. Different strategies are developed by participants to manoeuvre in the videoconferencing classrooms. In conclusion the four studies show how actor-networks influence the practice of local learning centres by representations.
Denna avhandling är en studie om lärcentra i Sverige. Syftet är att beskriva och öka kunskapen om de relationer och de aktörsnätverk som omger praktiken. Avhandlingen bygger på resultat från fyra olika delstudier. Fokus i två av dessa ligger på lärcentra som organisation. Hur ser relationen mellan lärcentra och omgivande aktörer ut i regionen och vilka strategier används för att skapa aktörsnätverk? De övriga två studierna handlar om videokonferens där fokus ligger på hur relationer skapas mellan miljö, teknik och människor. Särskilt studeras interaktion och kommunikation mellan dessa aktörer i en utpräglat socioteknisk lärandemiljö. Den teoretiska ramen för de olika delstudierna är aktörsnätverksteori som används tillsammans med begrepp som spatiala relationer. De fyra studierna använder sig i huvudsak av kvalitativa metoder som intervjuer och observationsstudier. Datainsamlingen berör en bred samling informanter som rektorer, lärare och studenter tillsammans med projektledare, politiker och företagare. Studiernas resultat visar att det finns skilda synsätt på utbildning och kompetens mellan olika grupper av aktörer. Trots involveringsstrategier av aktörer från omgivande aktörsnätverk lyckas man inte agera som en mäklare på en utbildningsmarknad. Resultaten visar vidare att miljö tillsammans med teknik har stort inflytande på hur studenter och lärare agerar i videokonferensklassrummen. Det är den materiella designen och den tekniska logiken som styr praktiken. Resultaten visar också på hur olika studerandestrategier utvecklas för att stå utanför interaktion i klassrummet tillsammans med hur lärares kommunikation utvecklas för att överbrygga avståndet till de studerande. Sammantaget visar de fyra studierna på hur olika aktörsnätverk inverkar på praktiken genom representationer.

Although there is limited research into the success of primary school networking initiatives in the UK, there seems to be an unquestioning faith displayed at national Government level for school collaborative working arrangements as a key means for driving forward whole school improvement. This research considers the possible benefits and challenges of one such initiative – Primary Strategy Learning Networks (DfES, 2004a). The research focuses on a reliance on school networks as power bases for promoting a national standards agenda. It considers the impact of an imposed model of school collaboration on the fluid nature of networking. It also acknowledges the benefits of a ‘network balance’ between the positive and negative features that impact on a network’s success and sustainability. Furthermore, the research explores the impact of power, authority and influence on the sustainability of networks. This is a qualitative study and data is gathered through interviews with network headteacher participants in two Primary Strategy Learning Networks over the course of an academic year. The research is also informed by an initial study of a Networked Learning Community (Hopkins and Jackson, 2002). Following an analysis of the findings, a number of recommendations are made. A suggested ‘ideal’ model for productive networking relationships among key stakeholders is offered for consideration and a Realistic Approach (Pawson, 2006) to evaluating such initiatives is argued to ensure a higher degree of success in implementing collaborative working practices for school improvement

The Spatial Reuse (SR) operation is gaining momentum in the latest IEEE 802.11 family of standards due to the overwhelming requirements posed by next-generation wireless networks. In particular, the rising traffic requirements and the number of concurrent devices compromise the efficiency of increasingly crowded Wireless Local Area Networks (WLANs) and throw into question their decentralized nature. The SR operation, initially introduced by the IEEE~802.11ax-2021 amendment and further studied in IEEE 802.11be-2024, aims to increase the number of concurrent transmissions in an Overlapping Basic Service Set (OBSS) using sensitivity adjustment and transmit power control, thus improving spectral efficiency. Our analysis of the SR operation shows outstanding potential in improving the number of concurrent transmissions in crowded deployments, which contributed to enabling low-latency next-generation applications. However, the potential gains of SR are currently limited by the rigidity of the mechanism introduced for the 11ax, and the lack of coordination among BSSs implementing it. The SR operation is evolving towards coordinated schemes where different BSSs cooperate. Nevertheless, coordination entails communication and synchronization overhead, which impact on the performance of WLANs remains unknown. Moreover, the coordinated approach is incompatible with devices using previous IEEE 802.11 versions, potentially leading to degrading the performance of legacy networks. For those reasons, in this thesis, we start assessing the viability of decentralized SR, and thoroughly examine the main impediments and shortcomings that may result from it. We aim to shed light on the future shape of WLANs concerning SR optimization and whether their decentralized nature should be kept, or it is preferable to evolve towards coordinated and centralized deployments. To address the SR problem in a decentralized manner, we focus on Artificial Intelligence (AI) and propose using a class of sequential learning-based methods, referred to as Multi-Armed Bandits (MABs). The MAB framework suits the SR problem because it addresses the uncertainty caused by the concurrent operation of multiple devices (i.e., multi-player setting) and the lack of information in decentralized deployments. MABs can potentially overcome the complexity of the spatial interactions that result from devices modifying their sensitivity and transmit power. In this regard, our results indicate significant performance gains (up to 100\% throughput improvement) in highly dense WLAN deployments. Nevertheless, the multi-agent setting raises several concerns that may compromise network devices' performance (definition of joint goals, time-horizon convergence, scalability aspects, or non-stationarity). Besides, our analysis of multi-agent SR encompasses an in-depth study of infrastructure aspects for next-generation AI-enabled networking.
L'operació de reutilització espacial (SR) està guanyant impuls per a la darrera família d'estàndards IEEE 802.11 a causa dels aclaparadors requisits que presenten les xarxes sense fils de nova generació. En particular, la creixent necessitat de tràfic i el nombre de dispositius concurrents comprometen l'eficiència de les xarxes d'àrea local sense fils (WLANs) cada cop més concorregudes i posen en dubte la seva naturalesa descentralitzada. L'operació SR, inicialment introduïda per l'estàndard IEEE 802.11ax-2021 i estudiada posteriorment a IEEE 802.11be-2024, pretén augmentar el nombre de transmissions concurrents en un conjunt bàsic de serveis superposats (OBSS) mitjançant l'ajustament de la sensibilitat i el control de potència de transmissió, millorant així l'eficiència espectral. El nostre estudi sobre el funcionament de SR mostra un potencial destacat per millorar el nombre de transmissions simultànies en desplegaments multitudinaris, contribuint així al desenvolupament d'aplicacions de nova generació de baixa latència. Tot i això, els beneficis potencials de SR són actualment limitats per la rigidesa del mecanisme introduït per a l'11ax, i la manca de coordinació entre els BSS que ho implementen. L'operació SR evoluciona cap a esquemes coordinats on cooperen diferents BSS. En canvi, la coordinació comporta una sobrecàrrega de comunicació i sincronització, el qual té un impacte en el rendiment de les WLAN. D'altra banda, l'esquema coordinat és incompatible amb els dispositius que utilitzen versions anteriors IEEE 802.11, la qual cosa podria deteriorar el rendiment de les xarxes ja existents. Per aquests motius, en aquesta tesi s'avalua la viabilitat de mecanismes descentralitzats per a SR i s'analitzen minuciosament els principals impediments i mancances que se'n poden derivar. El nostre objectiu és donar llum a la futura forma de les WLAN pel que fa a l?optimització de SR i si s'ha de mantenir el seu caràcter descentralitzat, o bé és preferible evolucionar cap a desplegaments coordinats i centralitzats. Per abordar SR de forma descentralitzada, ens centrem en la Intel·ligència Artificial (AI) i ens proposem utilitzar una classe de mètodes seqüencials basats en l'aprenentatge, anomenats Multi-Armed Bandits (MAB). L'esquema MAB s'adapta al problema descentralitzat de SR perquè aborda la incertesa causada pel funcionament simultani de diversos dispositius (és a dir, un entorn multi-jugador) i la falta d'informació que se'n deriva. Els MAB poden fer front a la complexitat darrera les interaccions espacials entre dispositius que resulten de modificar la seva sensibilitat i potència de transmissió. En aquest sentit, els nostres resultats indiquen guanys importants de rendiment (fins al 100 \%) en desplegaments altament densos. Tot i això, l'aplicació d'aprenentatge automàtic amb múltiples agents planteja diversos problemes que poden comprometre el rendiment dels dispositius d'una xarxa (definició d'objectius conjunts, horitzó de convergència, aspectes d'escalabilitat o manca d'estacionarietat). A més, el nostre estudi d'aprenentatge multi-agent per a SR multi-agent inclou aspectes d'infraestructura per a xarxes de nova generació que integrin AI de manera intrínseca.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2014.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 173-177).
We study decision making by networked entities and the interplay between networks and outcomes under two different contexts: in the first part of the thesis, we study how strategic agents that share local information coordinate; in the second part of the thesis, we quantify the gain of having access to temporally richer data for learning of influence networks. In the first part of the thesis, we study the role of local information channels in enabling coordination among strategic agents. Building on the standard finite-player global games framework, we show that the set of equilibria of a coordination game is highly sensitive to how information is locally shared among agents. In particular, we show that the coordination game has multiple equilibria if there exists a collection of agents such that (i) they do not share a common signal with any agent outside of that collection; and (ii) their information sets form an increasing sequence of nested sets, referred to as a filtration. Our characterization thus extends the results on the uniqueness and multiplicity of equilibria in global games beyond the well-known case in which agents have access to purely private or public signals. We then provide a characterization of how the extent of equilibrium multiplicity is determined by the extent to which subsets of agents have access to common information: we show that the size of the set of equilibrium strategies is increasing with the extent of variability in the size of the subsets of agents who observe the same signal. We study the set of equilibria in large coordination games, showing that as the number of agents grows, the game exhibits multiple equilibria if and only if a non-vanishing fraction of the agents have access to the same signal. We finally consider an application of our framework in which the noisy signals are interpreted to be the idiosyncratic signals of the agents, which are exchanged through a communication network. In the second part of the thesis, we quantify the gain in the speed of learning of parametric models of influence, due to having access to richer temporal information. We infer local influence relations between networked entities from data on outcomes and assess the value of temporal data by characterizing the speed of learning under three different types of available data: knowing the set of entities who take a particular action; knowing the order in which the entities take an action; and knowing the times of the actions. We propose a parametric model of influence which captures directed pairwise interactions and formulate different variations of the learning problem. We use the Fisher information, the Kullback-Leibler (KL) divergence, and sample complexity as measures for the speed of learning. We provide theoretical guarantees on the sample complexity for correct learning based on sets, sequences, and times. The asymptotic gain of having access to richer temporal data for the speed of learning is thus quantified in terms of the gap between the derived asymptotic requirements under different data modes. We also evaluate the practical value of learning with richer temporal data, by comparing learning with sets, sequences, and times given actual observational data. Experiments on both synthetic and real data, including data on mobile app installation behavior, and EEG data from epileptic seizure events, quantify the improvement in learning due to richer temporal data, and show that the proposed methodology recovers the underlying network well.
by Spyridon Ilias Zoumpoulis.
Ph. D.

During the last twenty years several models have been proposed to improve the classic Black-Scholes framework for equity derivatives pricing. Recently a new model has been proposed: Local-Stochastic Volatility Model (LSV). This model considers volatility as the product between a deterministic and a stochastic term. So far, the model choice was not only driven by the capacity of capturing empirically observed market features well, but also by the computational tractability of the calibration process. This is now undergoing a big change since machine learning technologies offer new perspectives on model calibration. In this thesis we consider the calibration problem to be the search for a model which generates given market prices and where additionally technology from generative adversarial networks can be used. This means parametrizing the model pool in a way which is accessible for machine learning techniques and interpreting the inverse problems a training task of a generative network, whose quality is assessed by an adversary. The calibration algorithm proposed for LSV models use as generative models so-called neural stochastic differential equations (SDE), which just means to parameterize the drift and volatility of an Ito-SDE by neural networks.

Mestrado Sistemas Sócio-Organizacionais da Actividade Económica
A discussão em torno das novas lógicas de desenvolvimento, fez emergir um crescente interesse pela perspectiva de um desenvolvimento assente numa lógica territorial de base regional e local. A esta dimensão, ressalta a capacidade de mobilização dos actores locais para a participação no processo de desenvolvimento das suas regiões e a ênfase na autonomia de decisão das comunidades territorialmente organizadas. Neste contexto, compete ao Estado, um papel fundamental de animador e catalisador da sociedade civil, tornando-se num parceiro activo em iniciativas locais. O presente trabalho pretende contribuir para a reflexão sobre a definição dos contornos de um modelo de intervenção na promoção do desenvolvimento, com base no exemplo extraído na experiência portuguesa das Redes Regionais para o Emprego (RRE) ao nível da mobilização, intervenção e integração de estratégias de desenvolvimento dos actores públicos e privados locais, cooperando num quadro de parceria. Tratando-se de uma experiência relativamente recente, cujo processo de constituição teve início em 1998, e não existindo informação sistematizada, as conclusões a que se chegou, decorrem dos resultados da análise empírica (efectuada a partir de um inquérito aplicado aos actores locais que integram as RRE) interpretados à luz do corpo teórico e ainda das considerações tecidas pelos entrevistados sobre esta problemática. Em síntese, apesar de confinadas a um contexto de actuação restrito, as RRE demonstram uma capacidade de concertação estratégica entre os actores locais em presença, mobilizados em torno de interesses comuns com base numa identidade territorial das respectivas regiões. São, contudo, fortemente pilotadas pelo Estado Central, e do ponto de vista da estrutura organizativa formal trata-se de uma articulação verticalizada, de lógica top-down cujo modelo de intervenção apesar de enquadrado inicialmente para a acção, veio a verificar uma oscilação para um modelo mais dirigido para a provisão.
The debate on the new logics of development has arisen an increasing interest in a developing perspective based on a territorial logic of regional and local support. To this extent, it stands out the local performers ability of mobilization by participating in the developing process of their regions, as well as the stress on the power of decision and autonomy of the communities organized by territorial areas. Within this context, it is also due to the State to deal with a fundamental matter, and become an active partner in local initiatives as a performer and promoter of civil society. The purpose of the current study is to make a contribution to the reflection on the definition of an intervention model to the promotion of the development, regarding the example given by the Portuguese experience within the Regional Networks for Employment (RNE) [Redes Regionais para o Emprego (RRE)], concerning the mobilization, intervention and integration through developing strategies of the local performers, both private and public, and cooperating in a partnership project. Considering this a rather new experience, whose process of constitution took place in the beginning of 1998, and since systematic information does not exist, we have reached to conclusions resulting from an empiric analysis (based on an inquiry applied to the local performers who belong to the RNE). The interpretation of the results was enlightened by the theoretical body, and took into account the comments on this matter made by the interviewees. In brief, despite the restricte context, the RRE show a capacity for strategic accord among the local agents, around common interests based in a territorial identity of the regions. They are, however, conduced by central State, and from the formal organizational structure, it's a verticalized connection suported in a top-down logic whose initial intervention action model was changed into a more directed model to provision.

Uno dei problemi più importanti della 3D Computer Vision è il cosiddetto surface matching, che consiste nel trovare corrispondenze tra oggetti tridimensionali. Attualmente il problema viene affrontato calcolando delle feature locali e compatte, chiamate descrittori, che devono essere riconosciute e messe in corrispondenza al mutare della posa dell'oggetto nello spazio, e devono quindi essere invarianti rispetto all'orientazione. Il metodo più usato per ottenere questa proprietà consiste nell'utilizzare dei Local Reference Frame (LRF): sistemi di coordinate locali che forniscono un'orientazione canonica alle porzioni di oggetti 3D che vengono usate per calcolare i descrittori. In letteratura esistono diversi modi per calcolare gli LRF, ma fanno tutti uso di algoritmi progettati manualmente. Vi è anche una recente proposta che utilizza reti neurali, tuttavia queste vengono addestrate mediante feature specificamente progettate per lo scopo, il che non permette di sfruttare pienamente i benefici delle moderne strategie di end-to-end learning. Lo scopo di questo lavoro è utilizzare un approccio data-driven per far imparare a una rete neurale il calcolo di un Local Reference Frame a partire da point cloud grezze, producendo quindi il primo esempio di end-to-end learning applicato alla stima di LRF. Per farlo, sfruttiamo una recente innovazione chiamata Spherical Convolutional Neural Networks, le quali generano e processano segnali nello spazio SO(3) e sono quindi naturalmente adatte a rappresentare e stimare orientazioni e LRF. Confrontiamo le prestazioni ottenute con quelle di metodi esistenti su benchmark standard, ottenendo risultati promettenti.

In this thesis, we focus on the so-called spectrum management's joint problem: efficient allocation of primary and secondary channels in channel bonding wireless local area networks (WLANs). From IEEE 802.11n to more recent standards like 802.11ax and 802.11be, bonding channels together is permitted to increase transmissions' bandwidth. While such an increase favors the potential network capacity and the activation of higher transmission rates, it comes at the price of reduced power per Hertz and accentuated issues on contention and interference with neighboring nodes. So, if WLANs were per se complex deployments, they are becoming even more complicated due to the increasing node density and the new technical features required by novel highly bandwidth-demanding applications. This dissertation provides an in-depth study of channel allocation and channel bonding in WLANs and discusses the suitability of solutions ranging from heuristic-based to reinforcement learning (RL)-based. To characterize channel bonding in saturated WLANs, we first propose an analytical model based on continuous-time Markov networks (CTMNs). This model relies on a novel, purpose-designed algorithm that generates CTMNs from spatially distributed scenarios, where nodes are not required to be within the carrier sense range of each other. We identify the key factors affecting the throughput and fairness of different channel bonding policies and expose critical interrelations among nodes in the spatial domain. By extending the analytical model to support unsaturated regimes, we highlight the benefits of allocating channels as wide as possible all together with adaptive policies to cope with unfair situations. Apart from the analytical model, this thesis relies on simulations to generalize channel bonding in dense scenarios while avoiding costly, sometimes unfeasible, experimental testbeds. Unfortunately, existing wireless network simulators tend to be too simplistic or too computational demanding. That is why we develop the Komondor wireless network simulator, with the essential advantage over other well-known simulators lying in its high event processing rate. We then deviate from analytical models and simulations and tackle real measurements through the Wi-Fi All-Channel Analyzer (WACA), the first system specifically designed to simultaneously measure the energy in all the 24 bondable Wi-Fi channels at the 5 GHz band. With WACA, we perform a first-of-its-kind spectrum measurement in areas including urban hotspots, residential neighborhoods, universities, and even a football match in Futbol Club Barcelona’s Camp Nou stadium. Our experimental findings reveal the underpinning factors controlling throughput gain, from which we highlight the inter-channel correlation. %We show the significance of the gathered dataset for finding new insights, which would not be possible otherwise, given that simple channel occupancy models severely underestimate the potential gains. As for solution proposals, we first cover heuristic-based approaches to find satisfactory configurations quickly. In this regard, we propose dynamic-wise (DyWi), a lightweight, decentralized, online primary channel selection algorithm for dynamic channel bonding. DyWi improves the expected WLAN throughput by considering not only the occupancy of the target primary channel but also the activity in the secondary channels. Even when assuming significant delays due to primary channel switching, simulations reveal important throughput and delay improvements. Finally, we identify machine learning (ML) approaches applicable to the spectrum management problem in WLANs and justify why model-free RL suits it the most. In particular, we put the focus on the adequate performance of stateless variations of RL and anticipate multi-armed bandits as the right solution since i) we need fast adaptability to suit user experience in dynamic Wi-Fi scenarios and ii) the number of multichannel configurations a network can adopt is limited; thus, agents can fully explore the action space in a reasonable time.
En aquesta tesi ens centrem en el problema conjunt de la gestió de l'espectre: assignació de canals primaris i secundaris a xarxes d'àrea local sense fils (WLAN) amb channel bonding. Des de l'estàndard IEEE 802.11n fins a estàndards més recents com el 802.11ac, el 802.11ax i el 802.11be, s'han anat proposant amplades de banda més grans per permetre agrupar canals, augmentant així l'amplada de banda total per transmissió. Tot i que aquest augment en l'amplada de banda afavoreix la capacitat potencial de les xarxes, suportant així els requeriments de les noves aplicacions Wi-Fi, també redueix la potència per Hertz i accentua els problemes de contenció i interferència entre nodes veïns. En resum, si les xarxes WLANs ja eren complexes per se, s'estan tornant encara més complexes a causa de l'augment de la densitat de nodes i de les noves prestacions incloses als darrers estàndards. Primer proposem un model analític basat en xarxes Markov en temps continu (CTMN) per caracteritzar channel bonding en WLANs saturades. Aquest model es basa en un nou algorisme que genera CTMNs a partir d'escenaris distribuïts espaialment, on no és necessari que els nodes estiguin dins del rang de contenció de la resta. Identifiquem els factors claus que afecten el rendiment i l'equitat de les diferents polítiques de channel bonding i mostrem l'existència d'interrelacions crítiques entre nodes en forma de reacció en cadena. D'això se'n desprèn que no hi ha una política channel bonding òptima única per a cada escenari. En ampliar el model analític per donar suport a règims no saturats, destaquem els avantatges d'assignar els canals tan amplis com sigui possible a les WLAN i implementar polítiques d'accés adaptatiu per fer front a les situacions que poden aparèixer tant en termes de rendiment com d'equitat. A part dels models analítics, aquesta tesi es basa en simulacions per generalitzar escenaris evitant costosos bancs de proves experimentals, de vegades inviables. Malauradament, els simuladors de xarxes sense fils existents solen ser massa simplistes o molt costosos computacionalment. És per això que desenvolupem el simulador de xarxes sense fils Komondor, concebut com una eina de codi obert accessible (llesta per utilitzar) per a la investigació de xarxes sense fils. L’avantatge essencial de Komondor respecte d’altres simuladors sense fils coneguts rau en la seva elevada velocitat de processament d’esdeveniments. A continuació ens desviem de models analítics i simulacions i abordem mesures reals a través del Wi-Fi All-Channel Analyzer (WACA), el primer sistema que mesura simultàniament l'energia de tots els 24 canals que permeten channel bonding a la banda Wi-Fi dels 5 GHz. Amb WACA, realitzem un estudi únic de localitzacions que inclouen nuclis urbans, barris residencials, universitats i fins i tot un partit a al Camp Nou, un estadi ple amb 98.000 aficionats i 12.000 connexions Wi-Fi simultànies. Les dades experimentals revelen els factors fonamentals que controlen el guany de rendiment, a partir dels quals ressaltem la correlació entre canals. També mostrem la importància del conjunt de dades recopilades per trobar nous factors claus, que d'una altra manera no seria possible, atès que els models d'ocupació de canals simples subestimen els guanys potencials. Pel que fa a solucions, primer discutim propostes basades en heurístiques per trobar configuracions satisfactòries ràpidament. En aquest sentit, proposem dinàmicament (DyWi), un algorisme de selecció de canal primari en línia, descentralitzat i eficient per xarxes channel bonding. DyWi millora el rendiment esperat tenint en compte no només l’ocupació del canal primari objectiu, sinó també l’activitat dels canals secundaris. Fins i tot quan suposem retards significatius a causa del canvi de canal primari, observem millores importants en termes de rendiment i retard. Finalment, identifiquem els enfocaments d’aprenentatge automàtic (o machine learning) aplicables al problema de la gestió de l’espectre a les WLAN i justifiquem per què l'aprenentatge del tipus reinforcement learning (RL) és el més adient. En particular, ens centrem en el rendiment adequat de les variacions d'RL sense estats i proposem multi-armed bandits com la solució adequada, ja que i) necessitem una adaptabilitat ràpida per millorar l’experiència d’usuari en escenaris Wi-Fi dinàmics i ii) el nombre de configuracions multicanal que una xarxa pot adoptar és limitat; per tant, els agents poden explorar completament l’espai d’acció en un temps raonable.

The overall aim of this thesis is to, through a case study of a local disaster organization, study the prerequisites for building functional Hastily Formed Networks in disasters and also how HFN:s can be linked to a proactive disaster organization.The study looks upon HFN:s from the organizational, cognitive aspects of communication – not the network/information technology – and focus on the issue of local empowerment. To set the theoretical framework the thesis makes use of Peter Senge's model on learning organization, Karl Weicks organizational and communication theory of microlevels in systems, commitment and enactment. The method is a case study with interviews and field observation of a proactive disaster organization in Indonesia, in the Sultanate of Yogyakarta on Java. The results shows that Yogyakarta is actively working with a model to take advantage of the experience from earlier catastrophes, spread the learnings to other, integrate community in the learning and trying to build local empowerment including a bottom-up approach. In the interviews topics raised as important to create a good collaboration in a HFN are: 1) Provide local leaders with professional support and coordination based on a common vision. 2) Establishing local empowerment and common visions. 3) Development of the organization and individuals. 4) The need for formal and informal structures and communication. 5) Need for local mandate for coordination and increased competence in local management. The themes all deal with challenges for HFN:s, presented in earlier research, and link strongly to Senge’s theoretical model for learning organization, trying to build trust and also micro committment for uncertain periods. This contributes to the capacity to be prepared for the unprepared.

In the literature there is a great emphasis on the theories of regional growth and development. In many theories, the role of small and medium sized enterprises (SMEs) in the economic growth and development of different regions has been widely discussed. Therefore, there is an emphasis upon the importance of the SMEs in the growth and development of the regional and national economies. The aim of this study is to clarify the role of local capabilities in the exporting SMEs and consequently define their role in regional growth and shed some light on the situation of the local capabilities in Ankara province. Therefore, in this study, firstly the theoretical framework of regional growth theories and the role of exporting SMEs in the regional economic growth are constructed. Secondly, the increasing role of different types of SMEs in the regional growth and their defining characteristics are discussed. Thirdly, brief remarks on SMEs in the Turkish economy and Ankara have been provided. Afterwards, the main hypothesis of this study tested through the results iv iv obtained from the survey that was done with SMEs in Ankara province and success factors of SMEs are drawn from the in-depth interviews. Finally, by making a general evaluation some policy implications have been drawn.

Our world is becoming increasingly complex, and is rapidly changingwith distances being reduced. Societies today are also in atransition from traditional production industries to increasingreliance on communication, consumption, services and experience. Asone of these emerging ‘new industries’, tourism is part of thismovement. Globalisation also makes further development of tourismpossible through, amongst other things, the spreading of languages;the development of low-cost carriers; international monetary systems;telecommunications and other technological innovations. Tourismcontributes to the globalisation of society and at the same time is aproduct of it. This is confirmed by complexity theories that stressthe systematic and dynamic nature of globalisation and theinterdependence of the global and the local. Hence, in this thesis,it is argued that acknowledging the complexity of tourism isnecessary for understanding tourism development, and more knowledgeabout tourism also leads to greater knowledge of our society.

The main aim of this thesis is to understand the complexity of localand regional tourist destination development, by exploring social andcultural factors that influence this development. In order to fulfilthis aim, analysis has been conducted at three different levels:places, projects and people. More specifically, by examining placesand how they develop as destinations; investigating the structure of,and processes within, groups and networks important for destinationdevelopment and by exploring the roles, resources and attitudes ofenterprising people who are seen as key for development. The thesisis based on a research project including three studies of tourismdestinations and projects in Northern Sweden and Scotland.

The social and cultural factors connected to tourism development arein turn part of learning processes, which in this thesis are seen asfundamental mechanisms for processes of development. A frameworkcalled ‘Learning Destinations’ is introduced that demonstrates howimportant social and cultural factors manifest themselves at each ofthe three levels: places, projects and people. History and heritageand rationales are the main cultural factors discussed, whilstinteraction and boundaries are prominent social factors found toinfluence tourism development. It is suggested that the framework of‘Learning Destinations’ may serve as a tool for understanding thecomplexity of local and regional tourism development.

Face à la crise environnementale et climatique, les pratiques basées sur la gestion des processus écologiques ouvrent un nouvel horizon pour l’agriculture. L’enjeu actuel de l’agroforesterie basée sur les associations d’arbres, de cultures et d’élevage, n’est plus limité à la seule production, mais touche aussi à la durabilité́ de cette production, à la résilience du milieu cultivé vis-à-vis des aléas climatiques, à la fourniture des services écosystémiques et à la création de nouveaux « terroirs ». Il s’agit de concrétiser sa multifonctionnalité́ et son inscription dans le long terme. Or, ces pratiques agroécologiques ne peuvent pas résulter d’une pure application de recettes techniques ; elles se développent avec les caractères propres de chaque milieu accompagné par chaque agriculteur. L’entrée par les connaissances naturalistes et agroécologiques des agriculteurs, et par ce que l’on nomme paysage, peut-elle devenir un vecteur de transmission des pratiques agroforestières essentiellement par effet de voisinage et par apprentissage collectif ? Une étude critique des modalités et des dispositifs d’apprentissage, de transmission et d’accompagnement des pratiques agroforestières du Bassin francilien est menée à travers une démarche ethnopaysagère et géographique et par la sociologie pragmatique. Les récits des trajectoires agroforestières et en agriculture biologique ou de conservation des sols sont retracés à travers l’enquête ethnographique auprès d’agriculteurs et la collecte documentaire de terrain (photographies, dessins d’agriculteurs, plans de projets). Une typologie des différentes formes de paysages agroforestiers (linéaires d’arbres intra-parcellaire, complantés parfois d’une strate arbustive, maillage de haies champêtres, etc.) et la gamme des savoirs écologiques et de diversification qui y sont associés, est proposée. En regard, la méfiance envers les arbres chez certains agriculteurs de conservation des sols est identifiée. Dans un deuxième temps, l’enquête sur les modalités d’accompagnement par les parties prenantes du développement territorial (agriculteurs, propriétaires fonciers, agents des parcs naturels régionaux et des communautés de communes, vulgarisateurs), éclaire les conflits et les alliances entre acteurs territoriaux. Les outils et les processus d’apprentissage collectif de cette université agroforestière du dehors sont mis en évidence à travers les chroniques des ateliers de collectifs d’agriculteurs et des projets de recherche-participative autour du patrimoine de semences et de ligneux adaptés localement (enquête écologique, index-botanique, lecture et design paysager, ateliers de projets et de taille, sélection participative). Ce travail propose une synthèse des freins (le temps long, le manque de références locales et de savoir-faire de gestion, l’arbre dans le bail rural) et des conditions de transmission des pratiques agroforestières (implication des pionniers dans un groupe de pratiques, approche par l’expérience, implication des acteurs territoriaux). Celles-ci s’articulent entre différentes échelles : de la parcelle à la plaine, jusqu’à la communauté d’acteurs du « grand paysage ». Enfin, cette thèse identifie un ensemble d’expériences patrimoniales du vivant qui, à travers les bénéfices agro-écosystémiques et les filières alimentaires issus de ces nouveaux paysages agricoles, pourrait devenir le socle d’un projet local garant de la durabilité environnementale, sociale et économique du territoire
Faced with the environmental and climatic crisis, practices based on the management of ecological processes are opening up a new horizon for agriculture. The current challenge of agroforestry, based on associations of trees, crops and/or animals, is no longer limited to production alone, but also affects its sustainability and in particular the provision of ecosystem services and the food resilience of territories. However, these agroecological practices cannot result from the application of technical recipes. Can farmers’ naturalistic and agroecological knowledge of landscapes, become a vector for the transmission of agroforestry practices mainly through neighborhood effect and collective learning ? A critical study of the methods and mechanisms of learning, transmission and support of agroforestry practices in the Paris Basin is carried out through a transdisciplinary, ethno-geographic and pragmatic approach. The stories of their agroecological trajectories are retraced through ethnographic investigation and documentary collection from the field (photography, drawing of farmers, project plan). A typology of the different forms of agroforestry landscapes (intra-plot tree lines, diversified hedgerows gridding, etc.) and the range of ecological and diversification knowledge associated with them is proposed. In contrast, the mistrust of trees among some soil conservation farmers is identified. Secondly, the survey on the methods of support by stakeholders in territorial development (farmers, landowners, agricultural development associations, agents of regional nature parks and communities of municipalities), sheds light on conflicts and alliances between actors. The tools and collective learning processes of this outside agroecological university are highlighted through the chronicles of farmers' collective workshops and participatory research projects around the heritage of locally adapted seeds and woody plants (ecological survey, botanical index, reading and landscape design, project and pruning workshops, participatory selection). This work offers a synthesis of the brakes (the long time, the lack of local references and management know-how, the tree in the rural lease). And the conditions of transmission of agroforestry practices (pioneering attitudes, empirical and transversal approaches, involvement of territorial actors) are articulated between different scales: from the plot to the plain, to the community of actors in the large landscape. Finally, this thesis identifies a set of updating of living heritage which, through the agro-ecosystem benefits and the local food systems resulting from these new agricultural landscapes; could become the basis of a local project guaranteeing the environmental, social and economic sustainability of the territory

碩士
大葉大學
資訊管理學系碩士班
97
Since the technology of wireless local networks were invented, the character of data transmitting without being restricted by any physical media can let people access and share network resources anytime. Although the wireless infrastructure brings much convenience to us, at the same time, a large number of information is secret exposed. Hence, hackers can wiretap the transmitted data and embezzle the resources of wireless facilities, so that intrusion detection systems become indispensable roles wireless local networks. However, current intrusion detection systems arise too much false-alarm and false-negative with the result that system administrators should deeply concern about whether the alarm is useless and notice the effectiveness of rule base. Therefore, our research addresses this issue by integrating fuzzy association rules into the linear discriminant analysis (LDA) to construct a hybrid intrusion detection scheme. By employing misuse detection, we implement LDA to find out the unknown attacks which were hidden among the regular behaviors, then trace their attack patterns and characteristics by using the fuzzy association rules, and further automatically add the rules into the rule base to achieve the goal of self-learning functionality, such that it can increase the detecting rate of intrusion detection system and enhance the updating efficiency of rule base. Finally, an intrusion detection system was implemented to demonstrate the feasibility of our proposed scheme.

by Evan, Fung-yu Young.
Thesis (M.Phil.)--Chinese University of Hong Kong, 1993.
Includes bibliographical references (leaves 161-166).
List of Figures --- p.vi
List of Tables --- p.vii
List of Graphs --- p.viii
Abstract --- p.ix
Chapter Part I --- Learning Algorithms
Chapter 1 --- Representing Time in Connectionist Models --- p.1
Chapter 1.1 --- Introduction --- p.1
Chapter 1.2 --- Temporal Sequences --- p.2
Chapter 1.2.1 --- Recognition Tasks --- p.2
Chapter 1.2.2 --- Reproduction Tasks --- p.3
Chapter 1.2.3 --- Generation Tasks --- p.4
Chapter 1.3 --- Discrete Time v.s. Continuous Time --- p.4
Chapter 1.4 --- Time Delay Neural Network (TDNN) --- p.4
Chapter 1.4.1 --- Delay Elements in the Connections --- p.5
Chapter 1.4.2 --- NETtalk: An Application of TDNN --- p.7
Chapter 1.4.3 --- Drawbacks of TDNN --- p.8
Chapter 1.5 --- Networks with Context Units --- p.8
Chapter 1.5.1 --- Jordan's Network --- p.9
Chapter 1.5.2 --- Elman's Network --- p.10
Chapter 1.5.3 --- Other Architectures --- p.14
Chapter 1.5.4 --- Drawbacks of Using Context Units --- p.15
Chapter 1.6 --- Recurrent Neural Networks --- p.16
Chapter 1.6.1 --- Hopfield Models --- p.17
Chapter 1.6.2 --- Fully Recurrent Neural Networks --- p.20
Chapter A. --- EXAMPLES OF USING RECURRENT NETWORKS --- p.22
Chapter 1.7 --- Our Objective --- p.25
Chapter 2 --- Learning Algorithms for Recurrent Neural Networks --- p.27
Chapter 2.1 --- Introduction --- p.27
Chapter 2.2 --- Gradient Descent Methods --- p.29
Chapter 2.2.1 --- Backpropagation Through Time (BPTT) --- p.29
Chapter 2.2.2 --- Real Time Recurrent Learning Rule (RTRL) --- p.30
Chapter A. --- RTRL WITH TEACHER FORCING --- p.32
Chapter B. --- TERMINAL TEACHER FORCING --- p.33
Chapter C. --- CONTINUOUS TIME RTRL --- p.33
Chapter 2.2.3 --- Variants of RTRL --- p.34
Chapter A. --- SUB GROUPED RTRL --- p.34
Chapter B. --- A FIXED SIZE STORAGE 0(n3) TIME COMPLEXITY LEARNGING RULE --- p.35
Chapter 2.3 --- Non-Gradient Descent Methods --- p.37
Chapter 2.3.1 --- Neural Bucket Brigade (NBB) --- p.37
Chapter 2.3.2 --- Temporal Driven Method (TO) --- p.38
Chapter 2.4 --- Comparison between Different Approaches --- p.39
Chapter 2.5 --- Conclusion --- p.41
Chapter 3 --- Locally Connected Recurrent Networks --- p.43
Chapter 3.1 --- Introduction --- p.43
Chapter 3.2 --- Locally Connected Recurrent Networks --- p.44
Chapter 3.2.1 --- Network Topology --- p.44
Chapter 3.2.2 --- Subgrouping --- p.46
Chapter 3.2.3 --- Learning Algorithm --- p.47
Chapter 3.2.4 --- Continuous Time Learning Algorithm --- p.50
Chapter 3.3 --- Analysis --- p.51
Chapter 3.3.1 --- Time Complexity --- p.51
Chapter 3.3.2 --- Space Complexity --- p.51
Chapter 3.3.3 --- Local Computations in Time and Space --- p.51
Chapter 3.4 --- Running on Parallel Architectures --- p.52
Chapter 3.4.1 --- Mapping the Algorithm to Parallel Architectures --- p.52
Chapter 3.4.2 --- Parallel Learning Algorithm --- p.53
Chapter 3.4.3 --- Analysis --- p.54
Chapter 3.5 --- Ring-Structured Recurrent Network (RRN) --- p.55
Chapter 3.6 --- Comparison between RRN and RTRL in Sequence Recognition --- p.55
Chapter 3.6.1 --- Training Sets and Testing Sequences --- p.56
Chapter 3.6.2 --- Comparison in Training Speed --- p.58
Chapter 3.6.3 --- Comparison in Recalling Power --- p.59
Chapter 3.7 --- Comparison between RRN and RTRL in Time Series Prediction --- p.59
Chapter 3.7.1 --- Comparison in Training Speed --- p.62
Chapter 3.7.2 --- Comparison in Predictive Power --- p.63
Chapter 3.8 --- Conclusion --- p.65
Chapter Part II --- Applications
Chapter 4 --- Sequence Recognition by Ring-Structured Recurrent Networks --- p.67
Chapter 4.1 --- Introduction --- p.67
Chapter 4.2 --- Related Works --- p.68
Chapter 4.2.1 --- Feedback Multilayer Perceptron (FMLP) --- p.68
Chapter 4.2.2 --- Back Propagation Unfolded Recurrent Rule (BURR) --- p.69
Chapter 4.3 --- Experimental Details --- p.71
Chapter 4.3.1 --- Network Architecture --- p.71
Chapter 4.3.2 --- Input/Output Representations --- p.72
Chapter 4.3.3 --- Training Phase --- p.73
Chapter 4.3.4 --- Recalling Phase --- p.73
Chapter 4.4 --- Experimental Results --- p.74
Chapter 4.4.1 --- Temporal Memorizing Power --- p.74
Chapter 4.4.2 --- Time Warping Performance --- p.80
Chapter 4.4.3 --- Fault Tolerance --- p.85
Chapter 4.4.4 --- Learning Rate --- p.87
Chapter 4.5 --- Time Delay --- p.88
Chapter 4.6 --- Conclusion --- p.91
Chapter 5 --- Time Series Prediction --- p.92
Chapter 5.1 --- Introduction --- p.92
Chapter 5.2 --- Modelling in Feedforward Networks --- p.93
Chapter 5.3 --- Methodology with Recurrent Networks --- p.94
Chapter 5.3.1 --- Network Structure --- p.94
Chapter 5.3.2 --- Model Building - Training --- p.95
Chapter 5.3.3 --- Model Diagnosis - Testing --- p.95
Chapter 5.4 --- Training Paradigms --- p.96
Chapter 5.4.1 --- A Quasiperiodic Series with White Noise --- p.96
Chapter 5.4.2 --- A Chaotic Series --- p.97
Chapter 5.4.3 --- Sunspots Numbers --- p.98
Chapter 5.4.4 --- Hang Seng Index --- p.99
Chapter 5.5 --- Experimental Results and Discussions --- p.99
Chapter 5.5.1 --- A Quasiperiodic Series with White Noise --- p.101
Chapter 5.5.2 --- Logistic Map --- p.103
Chapter 5.5.3 --- Sunspots Numbers --- p.105
Chapter 5.5.4 --- Hang Seng Index --- p.109
Chapter 5.6 --- Conclusion --- p.112
Chapter 6 --- Chaos in Recurrent Networks --- p.114
Chapter 6.1 --- Introduction --- p.114
Chapter 6.2 --- Important Features of Chaos --- p.115
Chapter 6.2.1 --- First Return Map --- p.115
Chapter 6.2.2 --- Long Term Unpredictability --- p.117
Chapter 6.2.3 --- Sensitivity to Initial Conditions (SIC) --- p.118
Chapter 6.2.4 --- Strange Attractor --- p.119
Chapter 6.3 --- Chaotic Behaviour in Recurrent Networks --- p.120
Chapter 6.3.1 --- Network Structure --- p.121
Chapter 6.3.2 --- Dynamics in Training --- p.121
Chapter 6.3.3 --- Dynamics in Testing --- p.122
Chapter 6.4 --- Experiments and Discussions --- p.123
Chapter 6.4.1 --- Henon Model --- p.123
Chapter 6.4.2 --- Lorenz Model --- p.127
Chapter 6.5 --- Conclusion --- p.134
Chapter 7 --- Conclusion --- p.135
Appendix A Series 1 Sine Function with White Noise --- p.137
Appendix B Series 2 Logistic Map --- p.138
Appendix C Series 3 Sunspots Numbers from 1700 to 1979 --- p.139
Appendix D A Quasiperiodic Series with White Noise --- p.141
Appendix E Hang Seng Daily Closing Index in 1991 --- p.142
Appendix F Network Model for the Quasiperiodic Series with White Noise --- p.143
Appendix G Network Model for the Logistic Map --- p.144
Appendix H Network Model for the Sunspots Numbers --- p.145
Appendix I Network Model for the Hang Seng Index --- p.146
Appendix J Henon Model --- p.147
Appendix K Network Model for the Henon Map --- p.150
Appendix L Lorenz Model --- p.151
Appendix M Network Model for the Lorenz Map --- p.159
Bibliography --- p.161

碩士
國立中興大學
電機工程學系所
96
A new method for real-time object detection by a Fuzzy Neural Network with Principal Component-based Support Vector learning (FNN-PCSV) is proposed in this thesis. FNN-PCSV is a fuzzy system that consists of Takagi-Sugeno-Kang (TSK) type fuzzy rules. The antecedent part of FNN-PCSV is generated via fuzzy clustering of the input data. The dimension of free parameter vector in the consequent part of FNN-PCSV is first reduced by the PCA. A linear support vector machine is then used to tune the consequent parameters on the principal component space to give the network better generalization performance. The object detection system consists of two stages. The first stage uses color histogram of the global color appearance of an object as detection feature for a FNN-PCSV classifier. To represent color information by histograms as accurately as possible, a non-uniform partition of color space is proposed. An efficient method for histogram extraction during the image scanning process is proposed for real-time implementation. The second stage uses geometry-dependent local color appearance as color feature for another FNN-PCSV classifier. Candidates generated in stage one are filtered in this stage to reduce the number of false alarms. To verify performance of the proposed method, experiments on detection of two specific objects are performed. For comparison, other types of detection methods and classifiers are also applied to the same detection task. Results show the proposed FNN-PCSV-based detection system achieves better results than compared methods.

碩士
靜宜大學
資訊工程學系
107
In the modern industry, machine health monitoring systems are critical to the modern manufacturing industry. In the machine health monitoring, due to the advanced sensor technologies, the data-driven approach is becoming the most popular way. However, how to deal with the noise data and realize the spatial and temporal correlations presented in the data is a challenge. Traditional research focuses on the use of feature extraction methods to judge normal tool and wear tool for tool state. However, while detecting the characteristics of the worn tool, the tool is likely to cause damage to the workpiece and machine will be downtime. It will have a huge burden on time and material costs. With the rapid development and application of deep learning in recent years, data feature methods can be used to predict tool wear. It provides instant visibility into tool wear. The study proposes a deep learning model called Holistic-Local LSTM (HLLSTM) to predict tool wear. HLLSTM learns short-term data characteristics by segmenting data into fragment data. HLLSTM learns short-term data features by segmenting data into fragment data and short-term data features classify into holistic training and local training to capture more implicit feature information. HLLSTM learns short-term data features by segment data. Short-term data features classify into overall training and local training to capture more implicit feature information to predict tool wear. The data of the study is from 2010 phm society conference data challenge. It contains multiple sensor parameters and tool wear values. The model through the supervised learning method can train to learn the sensor parameters to predict the condition of tool wear. Finally, we compare the prediction error values of the HLLSTM model with the currently used neural network model. The model verifies the prediction accuracy by error values. The HLLSTM model presented in this paper predicts more accurate tool wear values. The HLLSTM model can reduce the mean absolute error between the actual and the predicted tool wear value by a factor of two compared to other neural network models.

Large scale semantic segmentation is considered as one of the fundamental tasks in 3D scene understanding. Point clouds provide a basic and rich geometric rep- resentation of scenes and tangible objects. Convolutional Neural Networks (CNNs) have demonstrated an impressive success in processing regular discrete data such as 2D images and 1D audio. However, CNNs do not directly generalize to point cloud processing due to their irregular and un-ordered nature. One way to extend CNNs to point cloud understanding is to derive an intermediate euclidean representation of a point cloud by projecting onto image domain, voxelizing, or treating points as vertices of an un-directed graph. Graph-CNNs (GCNs) have demonstrated to be a very promising solution for deep learning on irregular data such as social networks, bi- ological systems, and recently point clouds. Early works in literature for graph based point networks relied on constructing dynamic graphs in the node feature space to define a convolution kernel. Later works constructed hierarchical static graphs in 3D space for an encoder-decoder framework inspired from image segmentation. This thesis takes a closer look at both dynamic and static graph neighborhoods of graph- based point networks for the task of semantic segmentation in order to: 1) discuss a potential cause for why going deep in dynamic GCNs does not necessarily lead to an improved performance, and 2) propose a new approach in treating points in a static graph neighborhood for an improved information aggregation. The proposed method leads to an efficient graph based 3D semantic segmentation network that is on par with current state-of-the-art methods on both indoor and outdoor scene semantic segmentation benchmarks such as S3DIS and Semantic3D.

Au cours de la dernière décennie, l'apprentissage profond est devenu une composante majeure de l'intelligence artificielle, ayant mené à une série d'avancées capitales dans une variété de domaines. L'un des piliers de l'apprentissage profond est l'optimisation de fonction de coût par l'algorithme du gradient stochastique (SGD). Traditionnellement en apprentissage profond, les réseaux de neurones sont des fonctions mathématiques différentiables, et les gradients requis pour l'algorithme SGD sont calculés par rétropropagation. Cependant, les architectures informatiques sur lesquelles ces réseaux de neurones sont implémentés et entraînés souffrent d’inefficacités en vitesse et en énergie, dues à la séparation de la mémoire et des calculs dans ces architectures. Pour résoudre ces problèmes, le neuromorphique vise à implementer les réseaux de neurones dans des architectures qui fusionnent mémoire et calculs, imitant plus fidèlement le cerveau. Dans cette thèse, nous soutenons que pour construire efficacement des réseaux de neurones dans des architectures neuromorphiques, il est nécessaire de repenser les algorithmes pour les implémenter et les entraîner. Nous présentons un cadre mathématique alternative, compatible lui aussi avec l’algorithme SGD, qui permet de concevoir des réseaux de neurones dans des substrats qui exploitent mieux les lois de la physique. Notre cadre mathématique s'applique à une très large classe de modèles, à savoir les systèmes dont l'état ou la dynamique sont décrits par des équations variationnelles. La procédure pour calculer les gradients de la fonction de coût dans de tels systèmes (qui dans de nombreux cas pratiques ne nécessite que de l'information locale pour chaque paramètre) est appelée “equilibrium propagation” (EqProp). Comme beaucoup de systèmes en physique et en ingénierie peuvent être décrits par des principes variationnels, notre cadre mathématique peut potentiellement s'appliquer à une grande variété de systèmes physiques, dont les applications vont au delà du neuromorphique et touchent divers champs d'ingénierie.
In the last decade, deep learning has become a major component of artificial intelligence, leading to a series of breakthroughs across a wide variety of domains. The workhorse of deep learning is the optimization of loss functions by stochastic gradient descent (SGD). Traditionally in deep learning, neural networks are differentiable mathematical functions, and the loss gradients required for SGD are computed with the backpropagation algorithm. However, the computer architectures on which these neural networks are implemented and trained suffer from speed and energy inefficiency issues, due to the separation of memory and processing in these architectures. To solve these problems, the field of neuromorphic computing aims at implementing neural networks on hardware architectures that merge memory and processing, just like brains do. In this thesis, we argue that building large, fast and efficient neural networks on neuromorphic architectures also requires rethinking the algorithms to implement and train them. We present an alternative mathematical framework, also compatible with SGD, which offers the possibility to design neural networks in substrates that directly exploit the laws of physics. Our framework applies to a very broad class of models, namely those whose state or dynamics are described by variational equations. This includes physical systems whose equilibrium state minimizes an energy function, and physical systems whose trajectory minimizes an action functional (principle of least action). We present a simple procedure to compute the loss gradients in such systems, called equilibrium propagation (EqProp), which requires solely locally available information for each trainable parameter. Since many models in physics and engineering can be described by variational principles, our framework has the potential to be applied to a broad variety of physical systems, whose applications extend to various fields of engineering, beyond neuromorphic computing.

Participatory development is believed to be an essential ingredient of the development process because it enables the people to influence the policy and decision-making process. It facilitates the designing of and enhances the implementation of plans and programmes or projects and that participation results in the development of ownership and belonging among the people themselves. This feeling in turn goes a long way to ensure the success of a programme or project. This study highlights some of the elements that led to the successful implementation of an income-generating project with women. The study also demonstrates that the people's own ideas are able to command adequate levels of participation, support and commitment to enhance their ability to meet their needs. The study further looks at the ownership of the project as very important for successful implementation and that projects or programmes should be facilitated within the means of the local participants to manage and control them. The elements mentioned in this study are an essential part of participatory development and enable the people to learn to take charge of their lives and solve their own problems. The study further looks at cultural and traditional values as an essential part of human growth, because local action taken by the women is very important in promoting overall success of the project.
Social work
M.A. (SS) (Mental Health)