The distributed and privacy sensitive nature of cellular networks make them strong candidates for optimization using Federated Learning, but this exposes them to a problem inherent to the learning paradigm: performance inequality due to heterogeneous client data distributions. The prevailing approach of enforcing uniform client performance ignores client-specific performance limitations due to different levels of irreducible uncertainty present in their data, resulting in deteriorated network performance. To address this issue, this thesis introduces two novel federated algorithms designed to enhance learning efficiency and ensure fairness in the presence of heteroscedastic noise, reflecting the distributive justice principles of utilitarianism and equality. Under these circumstances, the proposed algorithms are shown to significantly improve overall performance and performance fairness. The deployment of these algorithms promises a dual benefit: enhancement in network performance and a fairer distribution of service quality for end users.

Decentralized Machine Learning could address some problematic facets with Federated Learning. There is no central server acting as an arbiter of whom or what may benefit from Machine Learning models created by the vast amount of data becoming available in recent years. It could also increase the reliability and scalability of Machine Learning systems thereby drawing the benefit of having more data accessible. Gossip Learning is such a protocol, but has primarily been designed with linear models in mind. How does Gossip Learning perform when training Deep Neural Networks? Could it be a viable alternative to Federated Learning? In this thesis, we implement Gossip Learning using two different model merging strategies. We also design and implement two extensions to this protocol with the goal of achieving higher performance when training under churn. The training methods are compared on two tasks: image classification on the Federated Extended MNIST dataset and time- series forecasting on the NN5 dataset. Additionally, we also run an experiment where learners churn, alternating between being available and unavailable. We find that Gossip Learning performs slightly better in settings where learners do not churn but is vastly outperformed in the setting where they do. / Decentraliserad Maskinginlärning kan lösa några problematiska aspekter med Federated Learning. Det finns ingen central server som agerar som domare för vilka som får gagna av Maskininlärningsmodellerna skapad av den stora mäng data som blivit tillgänglig på senare år. Det skulle också kunna öka pålitligheten och skalbarheten av Maskininlärningssystem och därav dra nytta av att mer data är tillgänglig. Gossip Learning är ett sånt protokoll, men det är primärt designat med linjära modeller i åtanke. Hur presterar Gossip Learning när man tränar Djupa Neurala Nätverk? Kan det vara ett möjligt alternativ till Federated Learning? I det här exjobbet implementerar vi Gossip Learning med två olika modelsammanslagningstekniker. Vi designar och implementerar även två tillägg till protokollet med målet att uppnå bättre prestanda när man tränar i system där noder går ner och kommer up. Träningsmetoderna jämförs på två uppgifter: bildklassificering på Federated Extended MNIST datauppsättningen och tidsserieprognostisering på NN5 datauppsättningen. Dessutom har vi även experiment då noder alternerar mellan att vara tillgängliga och otillgängliga. Vi finner att Gossip Learning presterar marginellt bättre i miljöer då noder alltid är tillgängliga men är kraftigt överträffade i miljöer då noder alternerar mellan att vara tillgängliga och otillgängliga.

Federated learning faces a challenge when dealing with highly heterogeneous data and it can sometimes be inadequate to adopt an approach where a single model is trained for usage at all nodes in the network. Different approaches have been investigated to succumb this issue such as adapting the trained model to each node and clustering the nodes in the network and train a different model for each cluster where the data is less heterogeneous. In this work we study the possibilities to improve the local model performance utilizing the hierarchical setup that comes with clustering the participating clients in the network. Experiments are carried out featuring a Long Short-Term Memory network to perform time series forecasting to evaluate different approaches utilizing the hierarchical setup and comparing them to standard federated learning approaches. The experiments are done using a dataset collected by Ericsson AB consisting of handovers recorded at base stations in an European city. The hierarchical approaches didn’t show any benefit over common two-level approaches. / Federated Learning står inför en utmaning när det gäller att hantera data med en hög grad av heterogenitet och det kan i vissa fall vara olämpligt att använda sig av en approach där en och samma modell är tränad för att användas av alla noder i nätverket. Olika approacher för att hantera detta problem har undersökts som att anpassa den tränade modellen till varje nod och att klustra noderna i nätverket och träna en egen modell för varje kluster inom vilket datan är mindre heterogen. I detta arbete studeras möjligheterna att förbättra prestandan hos de lokala modellerna genom att dra nytta av den hierarkiska anordning som uppstår när de deltagande noderna i nätverket grupperas i kluster. Experiment är utförda med ett Long Short-Term Memory-nätverk för att utföra tidsserieprognoser för att utvärdera olika approacher som drar nytta av den hierarkiska anordningen och jämför dem med vanliga federated learning-approacher. Experimenten är utförda med ett dataset insamlat av Ericsson AB. Det består av "handoversfrån basstationer i en europeisk stad. De hierarkiska approacherna visade inga fördelar jämfört med de vanliga två-nivåapproacherna.

A produção científica de uma área do conhecimento aparece em diferentes formatos e é disponibilizada de forma essencialmente distribuída por entre revistas, anais, teses, dissertações e outros formatos característicos utilizados pela comunidade científica para a sistematização de seu discurso. Uma federação de bibliotecas digitais oferece uma arquitetura da informação que tem por finalidade facilitar a agregação de diferentes tipos de documentos disponibilizados, facilitando termos acesso a esses documentos, bem como a seus metadados descritores, formando, desse modo, verdadeiras estruturas de apoio ao desenvolvimento de pesquisas e análises científicas dos documentos que por ali circulam. Já a análise de redes sociais vem se mostrando um importante objeto de pesquisa da área da Ciência da Informação nas últimas décadas, tendo sido apropriada ainda de forma preliminar pela comunidade científica brasileira. Como forma de ampliar o conhecimento e experimentações com o uso da análise de redes sociais e identificar seu potencial analítico em relação ao que poderíamos coletar de informações de uma federação de bibliotecas digitais, tivemos por objetivo neste trabalho utilizar a análise de rede para mapear os padrões, tendências e estratégias de conectividade de dois planos de relacionamento entre pesquisadores: a coautoria em documentos oriundos de revistas científicas e a participação em bancas de defesas de teses e dissertações. Além disso, buscamos mapear as causas sociais e políticas dos padrões de rede identificados, colocando em evidência um uso crítico e contextualizado dos indicadores estruturais e dinâmicos de redes utilizados neste trabalho. Utilizamos como caso a biblioteca digital federada Univerciencia.org, uma biblioteca especializada na área de Ciências da Comunicação, tendo fornecido como fonte de dados 49 revistas científicas da área com 9864 documentos e 12 bibliotecas digitais de teses e dissertações com 1961 documentos. Os resultados apontam que os movimentos geradores e constituintes das redes sociais em nossos dois planos de análise são fortemente determinados por uma racionalidade característica da política científica do campo da Comunicação e da ciência de modo geral. / The scientific production of an area of knowledge appears in different formats and is available in a distributed mainly through journals, proceedings, theses, dissertations and other typical formats used by the scientific community for the systematization of his speech. A federation of digital libraries offers an information architecture that aims to facilitate the aggregation of different types of documents available, facilitating access to those documents and their metadata descriptors, forming thus real structures to support the development of research and analysis of scientific documents that circulate through there. The analysis of social networks has proven an important subject of research in the area of Information Science and in recent decades have been appropriate even in a preliminary way by the Brazilian scientific community. As a way to increase knowledge and experimentation with the use of social network analysis and identify his potential analytical, the objective of this work was use network analysis to map the patterns , trends and connectivity strategies between two planes of relation between researchers: co-authoring of documents from scientific journals and participation in defenses of theses and dissertations. Furthermore, we seek to map the social and political causes of network patterns identified, highlighting a critical use of structural and dynamic indicators. We use as case Univerciencia.org federated digital library, a library specialized in the field of Communication Sciences and provided as a source of data collected 49 scientific journals in the area with 9864 documents and 12 digital libraries of theses and dissertations with 1961 documents. The results show that the generative movements and constituents of social networks in our two levels of analysis are strongly determined by a characteristic rationality of science policy in the field of communication and science in general.

Over the past thirty years, clinical research has benefited substantially from the adoption of electronic medical record systems. As deployment has increased, so too has the number of researchers seeking to improve the overall analytical environment by way of tools and models. Although much work has been done, there are still many uninvestigated areas; two of which are explored in this dissertation. The first pertains to the physical storage of the data itself. There are two generally accepted storage models: relational and entity-attribute-value (EAV). For clinical data, EAV systems are preferred due to their natural way of managing many-to-many relationships, sparse attributes, and dynamic processes along with minimal conversion effort and reduction in federation complexities. However, the relational database management systems on which they are implemented, are not intended to organize and retrieve data in this format; eroding their performance gains. To combat this effect, we present the foundation for an EAV Database Management System (EDBMS). We discuss data conversion methodologies, formulate the requisite metadata and partitioned type-sensing index structures, and provide detailed runtime and experimental analysis with five extant methods. Our results show that the prototype, EAVDB, reduces space and conversion requirements while enhancing overall query performance. The second topic concerns query performance in a federated environment. One method used to decrease query execution time, is to pre-compute and store "beneficial" queries (views). The View Selection Problem (VSP) identifies these views subject to resource constraints. A federated model, however, has yet to be developed. In this dissertation, we submit three advances in view materialization. First, a more robust optimization function, the Minimum-Maintenance View Selection Problem (MMVSP), is derived by combining existing approaches. Second, the Federated View Selection Problem (FVSP), built upon the MMVSP, and federated data cube lattice are formalized. The FVSP allows for multiple querying nodes, partial and full materialization, and data propagation constriction. The latter two are shown to greatly reduce the overall number of valid solutions within the solution space and thus a novel, multi-tiered approach is given. Lastly, EAV materialization, which is introduced in this dissertation, is incorporated into an expanded, multi-modal variant of the FVSP. As models and heuristics for both the federated and EAV VSP, to the best of our knowledge, do not exist, this research defines two new branches of data warehouse optimization. Coupled with our EDBMS design, this dissertation confronts two main challenges associated with clinical data warehousing and federation.

In order to remain competitive, the integration of their information systems is an imperative for many large organisations. Applications that originally have been developed independently are now required to interoperate to support new or different functions of the enterprise. Although the mechanisms for application interoperation exist provided by the technology, due to the sheer number and complexity of the running systems, integration solutions � centralised or distributed�appropriate at the local level do not translate successfully to the whole enterprise. Centralised integration approaches often satisfy only some of the integration requirements, they are very expensive, and are fraught with danger since they imply an �all or nothing� approach. Distributed approaches, on the other hand, suffer from complexity and scalability problems as the number of system interfaces to be implemented and the number of execution-time invocations grows with the number of component applications. This dissertation makes a contribution to the field of Enterprise Application Integration (EAI) within the framework of distributed systems technology. Based on real-life case studies experience, we present here a federated approach that controls the size and complexity of the integration effort by reusing existing systems as much as possible and reducing the number of interacting applications. Only selected local elements are exposed to the organisational milieu, and a consistent supporting infrastructure is provided to make systems interactions possible. Our approach provides a flexible and scalable strategy to enterprise integration, avoiding the shortcomings of traditional approaches. We respect existing organisational structures, and demonstrate how appropriate federation infrastructure and protocols enable the interoperation of existing systems. The three main facets of enterprise knowledge are systematically incorporated into the integration effort: a) by the use of domain ontologies to support data integration; b) by the development of a methodology to include business rules; and c) by the development of FEW, a federated workflow model to implement the business processes of the organisation.

This thesis addresses the issues of coherent distributed knowledge production and sharing in the Life-science area. In spite of the continuously increasing computing and storage capabilities of computing infrastructures, the management of massive scientific data through centralized approaches became inappropriate, for several reasons: (i) they do not guarantee the autonomy property of data providers, constrained, for either ethical or legal concerns, to keep the control over the data they host, (ii) they do not scale and adapt to the massive scientific data produced through e-Science platforms. In the context of the NeuroLOG and VIP Life-science collaborative platforms, we address on one hand, distribution and heterogeneity issues underlying, possibly sensitive, resource sharing ; and on the other hand, automated knowledge production through the usage of these e-Science platforms, to ease the exploitation of the massively produced scientific data. We rely on an ontological approach for knowledge modeling and propose, based on Semantic Web technologies, to (i) extend these platforms with efficient, static and dynamic, transparent federated semantic querying strategies, and (ii) to extend their data processing environment, from both provenance information captured at run-time and domain-specific inference rules, to automate the semantic annotation of ''in silico'' experiment results. The results of this thesis have been evaluated on the Grid'5000 distributed and controlled infrastructure. They contribute to addressing three of the main challenging issues faced in the area of computational science platforms through (i) a model for secured collaborations and a distributed access control strategy allowing for the setup of multi-centric studies while still considering competitive activities, (ii) semantic experiment summaries, meaningful from the end-user perspective, aimed at easing the navigation into massive scientific data resulting from large-scale experimental campaigns, and (iii) efficient distributed querying and reasoning strategies, relying on Semantic Web standards, aimed at sharing capitalized knowledge and providing connectivity towards the Web of Linked Data.

In the 1960s, Irene Spry served as the Federated Women's Institutes of Canada (FWIC) representative to the Associated Country Women of the World (ACWW). In 1967 she accepted an offer to be the ACWW deputy president, a post that she held until the mid-1970s. During this time, the ACWW and its member societies engaged in international development efforts around the world. This was a critical moment in the history of international development. The Canadian movement for development was propelled by domestic and global politics, as well as a changing society that embraced a sense of global citizenship. Arising out of this context and armoured with her own socialist politics, Spry carefully navigated the development efforts of the ACWW. These efforts straddled grassroots ideals and mainstream pressures from the United Nations (UN). As a women's Non-Governmental Organization (NGO), the ACWW was part of the initial force behind the global shift in the approach to development referred to as Women in Development (WID). Contemporary research, however, suggests that WID has not succeeded in addressing the concerns of women in "developing" countries. As a case study, this paper examines some of the historical roots of WID and identifies the historical continuities that persist in today's development discourse. Analyzing Spry's documents from the Library and Archives Canada through the lens of feminist postcolonial theory reveals the dominance of Eurocentric ideologies within the development practices of the ACWW. The impetus to reach out to help people in developing countries became socially and politically part of the Canadian identity and, as Spry's navigation through the discourses of the international agencies and ACWW members reveal, such sentiments of international benevolence were inherently neo-colonial. In much the same way that Himani Bannerji suggests that subjects are "invented," women involved in this movement intersected discourses of modernity and "race" with essentializing notions of gender, which contributed to a standardized practice of development. This case study ultimately demonstrates that good intentions were not enough to decolonize western women's efforts to "develop" parts of Africa in the late 1960s and early 1970s.

Stobbe, Mark Jacob, M.A. University of Saskatchewan, Saskatoon, Canada, February 2014. Public-Private Partnerships in Saskatchewan: A Tale of two Upgraders. Most of the literature dealing with public-private partnerships (P3s) examines the impact of private sector involvement in the provision of infrastructure or services normally provided by the public sector. This thesis uses the two case studies of the NewGrade Heavy Oil Upgrader and the Bi-Provincial Heavy Oil upgrader to examine the dynamics of P3s entered into by government in a market-driven, commercial sector for the purposes of promoting economic development. In the 1980’s, there was a political consensus in the Saskatchewan legislature that the province needed upgrading capacity to convert heavy crude oil into more marketable and valuable light synthetic crude and that the upgraders should be built through P3s. The result was the creation of the NewGrade and Bi-Provincial Upgraders. In the 1990’s, financial losses at both upgraders caused the Saskatchewan government to demand renegotiation of these partnerships. The thesis examines these partnerships in their initial negotiation, construction/operation and renegotiation in order to determine what environmental factors and internal dynamics contributed to the success or failure of the partnerships and the relations between the partners. The thesis argues that the upgraders successfully achieved their public policy objectives and gained the benefits of synergies arising from the differences between the public and private sector. However, the partnerships came under severe stress arising from a prolonged downturn in oil markets and the price of crude oil. The resulting financial losses caused the Saskatchewan government to seek a renegotiation of the terms of partnership. Despite this common cause of stress in the partnerships, the renegotiations of the agreements varied greatly. It is demonstrated that these differences arose from the financial structure of each partnership, the nature of the private sector partners and the number of partners involved in the project. The thesis provides some observations of potential value for governments and corporations considering entering partnerships for economic development projects. The differences between partners that can create synergistic benefits can also be the basis for the erosion of trust between the partners. The different financial tools used by government to participate in P3s can have significant impacts on both project viability and relationships between the partners.

Methods: We address these issues by transforming the TCGA data into the Semantic Web standard Resource Description Format (RDF), link it to relevant datasets in the Linked Open Data (LOD) cloud and further propose an efficient data distribution strategy to host the resulting 20.4 billion triples data via several SPARQL endpoints. Having the TCGA data distributed across multiple SPARQL endpoints, we enable biomedical scientists to query and retrieve information from these SPARQL endpoints by proposing a TCGA tailored federated SPARQL query processing engine named TopFed. Results: We compare TopFed with a well established federation engine FedX in terms of source selection and query execution time by using 10 different federated SPARQL queries with varying requirements. Our evaluation results show that TopFed selects on average less than half of the sources (with 100% recall) with query execution time equal to one third to that of FedX. Conclusion: With TopFed, we aim to offer biomedical scientists a single-point-of-access through which distributed TCGA data can be accessed in unison. We believe the proposed system can greatly help researchers in the biomedical domain to carry out their research effectively with TCGA as the amount and diversity of data exceeds the ability of local resources to handle its retrieval and parsing.