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.

The use of a mobile charger (MC) is a popular method to restore energy in wireless rechargeable sensor networks(WRSN), whose effectiveness depends critically on the recharging strategy employed by the MC. In this thesis, we propose a novel on-line recharging mechanism strategy, called Continuous Local Learning (CLL), which predicts the current energy level of the sensor nodes and dynamically updates the schedule to visit the nodes before their batteries get depleted. The strategy is based on simple computations done by the MC with little memory requirements, and the communication is strictly local (between the MC and neighbouring nodes). In spite of its simplicity, this strategy was experimentally shown to be highly effective in keeping the network perpetually operating, ensuring that the number of sensing holes (i.e., non-operational sensors due to battery depletion) and their duration are very small at any time, and achieving immortality (i.e., no node ever becoming nonoperational) under many settings even in large networks. We also studied the flexibility of CLL under a variety of network parameters, showing its applicability in various contexts. We particularly focused on network size, data rate, sensors’ battery-capacity, and speed of the MC, and studied their impact on operational size and disconnection time under a wide range of values. The experiments indicate the fact that the effectiveness of CLL holds under all considered settings. We then compared the proposed solution with the popular class of static strategies since they share with CLL the features of simplicity, strict local communication and small memory and computational requirements. Experimental results showed that CLL outperforms these strategies in effectiveness. Not only is the number of sensors that are operational at any time higher under CLL, but the average duration of a sensing hole is also significantly lower. Finally, we studied the adaptability of CLL to a network’s sudden changes, in particular changes in data rate, which we call spikes. We studied the impact of spikes parameters on the performance of CLL. Experimental results showed that CLL is capable of reacting and adapting to these sudden changes with only a slight increase in non-operational size and disconnection time.

No description available.

Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2018-04-17T17:50:53Z No. of bitstreams: 1 jessicafaciroli.pdf: 4444017 bytes, checksum: abe3e6ddfe30914227e7d32dab33489c (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2018-04-20T12:54:53Z (GMT) No. of bitstreams: 1 jessicafaciroli.pdf: 4444017 bytes, checksum: abe3e6ddfe30914227e7d32dab33489c (MD5) / Made available in DSpace on 2018-04-20T12:54:53Z (GMT). No. of bitstreams: 1 jessicafaciroli.pdf: 4444017 bytes, checksum: abe3e6ddfe30914227e7d32dab33489c (MD5) Previous issue date: 2018-02-16 / Diversos estudos têm evidenciado a importância dos efeitos da social network na participação de programas de transferência de renda. Nesse contexto, esse estudo tem por objetivo analisar empiricamente os efeitos da social network na probabilidade de uma família participar do Programa Bolsa Família (PBF). A network foi construída com base no estudo de Aizer & Currie (2004), usando famílias beneficiárias e não beneficiárias do PBF, que vivem no mesmo Código de Endereço Postal (CEP) e que são do mesmo grupo racial. O mecanismo atrelado ao efeito que se deseja mensurar é o de que essa social network pode ser determinante para que as famílias não beneficiárias do PBF aprendam sobre os critérios, elegibilidades e condicionalidades do programa com seus vizinhos do mesmo grupo racial, por meio do compartilhamento das informações. Como estratégia empírica, utilizou-se o método Logit com efeitos fixos, sendo as informações das famílias extraídas no Cadastro Único para os anos de 2013 até 2015. Os principais resultados obtidos evidenciam que a social network impacta positivamente na probabilidade de participação da família no PBF. Mesmo quando se controlam as características não observadas das famílias e a interação com a disponibilidade de contatos, o efeito da social network permanece positivo e significativo. Esses resultados sugerem a existência de uma aprendizagem local no PBF, em que os não beneficiários que vivem em áreas onde têm muitos beneficiários serão mais propensos a se informarem e, consequentemente, se tornarem beneficiários do PBF. / Several studies have evidenced the importance of the social network's effects on the participation of income transfer programs. In this context, this study aims to empirically analyze the effects of the social network on the probability of a family participating in the Bolsa Familia Program (PBF). The network was constructed based on the study of Aizer & Currie (2004), using beneficiaries and non-beneficiaries families of the PBF, who live in the same postal address code (CEP) and are of the same racial group. The mechanism linked to the effect to be measured is that this social network can be determinant for the non-beneficiaries families of the PBF to learn about the program's criteria, eligibilities and conditionalities with its neighbors of the same racial group, by sharing the information. As an empirical strategy, the Logit Method with fixed effects was used, with the information of the families extracted in the “Cadastro Único” for the years 2013 to 2015. The main results obtained show that the social network has a positive impact on the probability of family participation in the PBF. Even when controlling the unobserved characteristics of the families and the interaction with the availability of contacts, the social network effect remains positive and significant. These results suggest that there is local learning in the PBF, where non-beneficiaries living in areas where they have many beneficiaries will be more likely to become aware of and thus become beneficiaries of the PBF.

Dans le domaine de l'apprentissage machine, les caractéristiques des données varient généralement dans l'espace des entrées : la distribution globale pourrait être multimodale et contenir des non-linéarités. Afin d'obtenir de bonnes performances, l'algorithme d'apprentissage devrait alors être capable de capturer et de s'adapter à ces changements. Même si les modèles linéaires ne parviennent pas à décrire des distributions complexes, ils sont réputés pour leur passage à l'échelle, en entraînement et en test, aux grands ensembles de données en termes de nombre d'exemples et de nombre de fonctionnalités. Plusieurs méthodes ont été proposées pour tirer parti du passage à l'échelle et de la simplicité des hypothèses linéaires afin de construire des modèles aux grandes capacités discriminatoires. Ces méthodes améliorent les modèles linéaires, dans le sens où elles renforcent leur expressivité grâce à différentes techniques. Cette thèse porte sur l'amélioration des approches d'apprentissage locales, une famille de techniques qui infère des modèles en capturant les caractéristiques locales de l'espace dans lequel les observations sont intégrées.L'hypothèse fondatrice de ces techniques est que le modèle appris doit se comporter de manière cohérente sur des exemples qui sont proches, ce qui implique que ses résultats doivent aussi changer de façon continue dans l'espace des entrées. La localité peut être définie sur la base de critères spatiaux (par exemple, la proximité en fonction d'une métrique choisie) ou d'autres relations fournies, telles que l'association à la même catégorie d'exemples ou un attribut commun. On sait que les approches locales d'apprentissage sont efficaces pour capturer des distributions complexes de données, évitant de recourir à la sélection d'un modèle spécifique pour la tâche. Cependant, les techniques de pointe souffrent de trois inconvénients majeurs :ils mémorisent facilement l'ensemble d'entraînement, ce qui se traduit par des performances médiocres sur de nouvelles données ; leurs prédictions manquent de continuité dans des endroits particuliers de l'espace ; elles évoluent mal avec la taille des ensembles des données. Les contributions de cette thèse examinent les problèmes susmentionnés dans deux directions : nous proposons d'introduire des informations secondaires dans la formulation du problème pour renforcer la continuité de la prédiction et atténuer le phénomène de la mémorisation ; nous fournissons une nouvelle représentation de l'ensemble de données qui tient compte de ses spécificités locales et améliore son évolutivité. Des études approfondies sont menées pour mettre en évidence l'efficacité de ces contributions pour confirmer le bien-fondé de leurs intuitions. Nous étudions empiriquement les performances des méthodes proposées tant sur des jeux de données synthétiques que sur des tâches réelles, en termes de précision et de temps d'exécution, et les comparons aux résultats de l'état de l'art. Nous analysons également nos approches d'un point de vue théorique, en étudiant leurs complexités de calcul et de mémoire et en dérivant des bornes de généralisation serrées. / In Machine Learning field, data characteristics usually vary over the space: the overall distribution might be multi-modal and contain non-linearities.In order to achieve good performance, the learning algorithm should then be able to capture and adapt to these changes. Even though linear models fail to describe complex distributions, they are renowned for their scalability, at training and at testing, to datasets big in terms of number of examples and of number of features. Several methods have been proposed to take advantage of the scalability and the simplicity of linear hypotheses to build models with great discriminatory capabilities. These methods empower linear models, in the sense that they enhance their expressive power through different techniques. This dissertation focuses on enhancing local learning approaches, a family of techniques that infers models by capturing the local characteristics of the space in which the observations are embedded. The founding assumption of these techniques is that the learned model should behave consistently on examples that are close, implying that its results should also change smoothly over the space. The locality can be defined on spatial criteria (e.g. closeness according to a selected metric) or other provided relations, such as the association to the same category of examples or a shared attribute. Local learning approaches are known to be effective in capturing complex distributions of the data, avoiding to resort to selecting a model specific for the task. However, state of the art techniques suffer from three major drawbacks: they easily memorize the training set, resulting in poor performance on unseen data; their predictions lack of smoothness in particular locations of the space;they scale poorly with the size of the datasets. The contributions of this dissertation investigate the aforementioned pitfalls in two directions: we propose to introduce side information in the problem formulation to enforce smoothness in prediction and attenuate the memorization phenomenon; we provide a new representation for the dataset which takes into account its local specificities and improves scalability. Thorough studies are conducted to highlight the effectiveness of the said contributions which confirmed the soundness of their intuitions. We empirically study the performance of the proposed methods both on toy and real tasks, in terms of accuracy and execution time, and compare it to state of the art results. We also analyze our approaches from a theoretical standpoint, by studying their computational and memory complexities and by deriving tight generalization bounds.

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.