Artificial Neural Networks have been widely probed by worldwide researchers to cope with the problems such as function approximation and data simulation. This thesis deals with Feed-forward Neural Networks (FNN's) with a new neuron activation function called Neuron-adaptive Activation Function (NAF), and Feed-forward Higher Order Neural Networks (HONN's) with this new neuron activation function. We have designed a new neural network model, the Neuron-Adaptive Neural Network (NANN), and mathematically proved that one NANN can approximate any piecewise continuous function to any desired accuracy. In the neural network literature only Zhang proved the universal approximation ability of FNN Group to any piecewise continuous function. Next, we have developed the approximation properties of Neuron Adaptive Higher Order Neural Networks (NAHONN's), a combination of HONN's and NAF, to any continuous function, functional and operator. Finally, we have created a software program called MASFinance which runs on the Solaris system for the approximation of continuous or discontinuous functions, and for the simulation of any continuous or discontinuous data (especially financial data). Our work distinguishes itself from previous work in the following ways: we use a new neuron-adaptive activation function, while the neuron activation functions in most existing work are all fixed and can't be tuned to adapt to different approximation problems; we only use on NANN to approximate any piecewise continuous function, while a neural network group must be utilised in previous research; we combine HONN's with NAF and investigate its approximation properties to any continuous function, functional, and operator; we present a new software program, MASFinance, for function approximation and data simulation. Experiments running MASFinance indicate that the proposed NANN's present several advantages over traditional neuron-fixed networks (such as greatly reduced network size, faster learning, and lessened simulation errors), and that the suggested NANN's can effectively approximate piecewise continuous functions better than neural networks groups. Experiments also indicate that NANN's are especially suitable for data simulation / Doctor of Philosophy (PhD)

Purpose - The purpose of this study is to examine and map out the network innovative companies and to calculate values of the network structure in order to compare them to different performance measures. Furthermore, we aim to investigate the trade-off efficiency of innovations in a particular network structure, more specifically to investigate if the same elements generating more innovations have a relationship with economic performance that originates from innovations.   Methodology - This study give emphasis to map and illustrate the Swedish companies on NASDAQ OMX First North network through direct and indirect connection and to compare the centrality, density and size of the companies ego network in our population with the performance measures which are logically connected with the launch of an innovation; average EBITDA (earnings before interest, taxes, depreciation and amortization) and average annual turnover. Findings - First we noticed that there was a significant connection between a negative average EBITDA and positive average annual turnover for our population, as we foretold would occur during the launch of an innovation. Secondly, the paper suggests that there is a weak or near non-existent connection between the elements that generates more innovations and the result of innovations, e.g. the economic performance of innovative firms. This might indicate that the focus of recent studies in the subject might have been mistaken focusing on the quantity of innovation, when the basic assumption of an innovation is that it is qualitative and thereby generates money for the company. This study suggests that more innovations do not necessarily lead to better economic performance for the companies within our population.

We develop two algorithms for important problems in mobile ad hoc networks (MANETs). A MANET is a collection of mobile processors (“nodes”) which communicate via message passing over wireless links. Each node can communicate directly with other nodes within a specified transmission radius; other communication is accomplished via message relay. Communication links may go up and down in a MANET (as nodes move toward or away from each other); thus, the MANET can consist of multiple connected components, and connected components can split and merge over time. We first present a deterministic leader election algorithm for asynchronous MANETs along with a correctness proof for it. Our work involves substantial modifications of an existing algorithm and its proof, and we adapt the existing algorithm to the asynchronous environment. Our algorithm’s running time and message complexity compare favorably with existing algorithms for leader election in MANETs. Second, many algorithms for MANETs require or can benefit from knowledge about the size of the network in terms of the number of processors. As such, we present an algorithm to approximately determine the size of a MANET. While the algorithm’s approximations of network size are only rough ones, the algorithm has the important qualities of requiring little communication overhead and being tolerant of link failures.

<p>Structured overlay networks form a major classof peer-to-peer systems, which are touted for their abilitiesto scale, tolerate failures, and self-manage. Any long livedInternet-scale distributed system is destined to facenetwork partitions. Although the problem of network partitionsand mergers is highly related to fault-tolerance andself-management in large-scale systems, it has hardly beenstudied in the context of structured peer-to-peer systems.These systems have mainly been studied under churn (frequentjoins/failures), which as a side effect solves the problemof network partitions, as it is similar to massive nodefailures. Yet, the crucial aspect of network mergers has beenignored. In fact, it has been claimed that ring-based structuredoverlay networks, which constitute the majority of thestructured overlays, are intrinsically ill-suited for mergingrings. In this thesis, we present a number of research papers representing our work on handling network partitions and mergers in structured overlay networks. The contribution of this thesis is threefold. First, we provide a solution for merging ring-based structured overlays. Our solution is tuneable, by a {\em fanout} parameter, to achieve a trade-off between message and time complexity. Second, we provide a network size estimation algorithm for ring-based structured overlays. We believe that an estimate of the current network size can be used for tuning overlay parameters that change according to the network size, for instance the fanout parameter in our merger solution.Third, we extend our work from fixing routing anomalies to achieving data consistency. We argue that decreasing lookup inconsistencies on the routing level aids in achieving data consistency in applications built on top of overlays. We study the frequency of occurence of lookup inconsistencies and discuss solutions to decrease the affect of lookup inconsistencies.</p>

<p>Business networks exist in many different situations the question of interest is; what makes these networks sustainable over a long period of time? This is a multiple case study with four network projects from the Swedish rural area of Storsjöbygden in Jämtland.</p><p>The Swedish rural areas consist of a strong majority of micro companies and small companies. To develop some of these areas the European Rural Development Program, LEADER helps funding projects for entrepreneurs and organisations with good ideas for development, education and cooperation activities.</p><p>This paper investigates four LEADER funded projects with a focus on business cooperation and networking. The projects were done between years 1995 and 2000. Some of the networks are still active today and others are not.  The aim here is to see if the three factors during the start up of the networks; natural and spontaneous relationships, project leader’s attitude and the number of participants, have an impact on the long term sustainability of these networks.</p><p>The empirical findings lead to the following conclusions; within the relationship factor time and attitude variables had an increased effect on the survival of the networks. The project leader’s knowledge, attitude and commitment also affected the outcome of the network projects. The third factor, the number of active members during the project had a great impact of the survival since all the networks lost members over time.</p>

Business networks exist in many different situations the question of interest is; what makes these networks sustainable over a long period of time? This is a multiple case study with four network projects from the Swedish rural area of Storsjöbygden in Jämtland. The Swedish rural areas consist of a strong majority of micro companies and small companies. To develop some of these areas the European Rural Development Program, LEADER helps funding projects for entrepreneurs and organisations with good ideas for development, education and cooperation activities. This paper investigates four LEADER funded projects with a focus on business cooperation and networking. The projects were done between years 1995 and 2000. Some of the networks are still active today and others are not.  The aim here is to see if the three factors during the start up of the networks; natural and spontaneous relationships, project leader’s attitude and the number of participants, have an impact on the long term sustainability of these networks. The empirical findings lead to the following conclusions; within the relationship factor time and attitude variables had an increased effect on the survival of the networks. The project leader’s knowledge, attitude and commitment also affected the outcome of the network projects. The third factor, the number of active members during the project had a great impact of the survival since all the networks lost members over time.

Structured overlay networks form a major classof peer-to-peer systems, which are touted for their abilitiesto scale, tolerate failures, and self-manage. Any long livedInternet-scale distributed system is destined to facenetwork partitions. Although the problem of network partitionsand mergers is highly related to fault-tolerance andself-management in large-scale systems, it has hardly beenstudied in the context of structured peer-to-peer systems.These systems have mainly been studied under churn (frequentjoins/failures), which as a side effect solves the problemof network partitions, as it is similar to massive nodefailures. Yet, the crucial aspect of network mergers has beenignored. In fact, it has been claimed that ring-based structuredoverlay networks, which constitute the majority of thestructured overlays, are intrinsically ill-suited for mergingrings. In this thesis, we present a number of research papers representing our work on handling network partitions and mergers in structured overlay networks. The contribution of this thesis is threefold. First, we provide a solution for merging ring-based structured overlays. Our solution is tuneable, by a {\em fanout} parameter, to achieve a trade-off between message and time complexity. Second, we provide a network size estimation algorithm for ring-based structured overlays. We believe that an estimate of the current network size can be used for tuning overlay parameters that change according to the network size, for instance the fanout parameter in our merger solution.Third, we extend our work from fixing routing anomalies to achieving data consistency. We argue that decreasing lookup inconsistencies on the routing level aids in achieving data consistency in applications built on top of overlays. We study the frequency of occurence of lookup inconsistencies and discuss solutions to decrease the affect of lookup inconsistencies.

<p>The capacity of deep reinforcement learning policy networks has been found to affect the performance of trained agents. It has been observed that policy networks with more parameters have better training performance and generalization ability than smaller networks. In this work, we find cases where this does not hold true. We observe unimodal variance in the zero-shot test return of varying width policies, which accompanies a drop in both train and test return. Empirically, we demonstrate mostly monotonically increasing performance or mostly optimal performance as the width of deep policy networks increase, except near the variance mode. Finally, we find a scenario where larger networks have increasing performance up to a point, then decreasing performance. We hypothesize that these observations align with the theory of double descent in supervised learning, although with specific differences.</p>

Methicillin-resistant Staphylococcus aureus (MRSA) is an antibiotic-resistant bacterium of epidemiologic importance in Canadian healthcare facilities. The contact between MRSA colonized or infected patients with other patients, healthcare workers (HCWs) and/or the healthcare environment can result in MRSA transmission and healthcare-associated MRSA (HA-MRSA) infections in hospitals. These HA-MRSA infections are linked with increased length of hospital stay, economic burden, morbidity and mortality. Although infection prevention and control programs initiated in 2009 in Canada and other developed countries (e.g., UK, France, Belgium, Denmark, etc.) have been relatively successful in reducing the rate of HA-MRSA infections, they continue to pose a threat to patients, especially to the more vulnerable in long term care and geriatric institutions. Historically, MRSA was a problem mainly in hospital settings but after mid-1990s new strains of MRSA have been identified among people without healthcare-related risks and have been classified as community-associated MRSA (CA-MRSA). Furthermore, the distinction between HA-MRSA and CA-MRSA strains is gradually waning due to both the introduction of HA-MRSA in communities, and the emergence of CA-MRSA strains in hospitals. The purpose of this thesis was to explore the feasibility of constructing healthcare networks to evaluate the role of healthcare providers (e.g., physicians) and places (e.g., patient rooms) in the transmission of MRSA in a large tertiary care facility. Method of investigation: a secondary data case-control study, using individual characteristics and network structure measures, conducted at The Ottawa Hospital (TOH) between April 1st, 2013 and March 31th, 2014. Results: It was feasible to build social networks in a large tertiary care facility based on electronic medical records data. The networks' size (represented by the number of vertices and lines) increased during the outbreak period (period 1) compared to the pre-outbreak period (period 0) for both groups and at all three TOH campuses. The calculated median degree centrality showed significant increase in value for both study groups during period 1 compared to period 0 for two of the TOH campuses (Civic and General). There was no significant difference between the median degree centrality calculated for each study group at the Heart Institute when compared for the two reference periods. The median degree centrality of the MRSA case group for period 0 showed no significant difference when compared to the same measure determined for the control group for all three TOH campuses. However, the median degree centrality calculated for period 1 was significantly increased for the control group compared to the MRSA case group for two TOH campuses (Civic and General) but showed no significant difference between the two groups from the Heart Institute. In addition, there was a correlation between the two network measures (degree centrality and eigenvector centrality) calculated to determine the most influential person or place in the MRSA case group networks. However, there was no correlation between the two network’s measures calculated for physicians included in MRSA case group networks. Conclusions: It is feasible to use social network analysis as an epidemiologic analysis tool to characterize the MRSA transmission in a hospital setting. The network's visible changes between the groups and reference periods were reflected by the network measures and supported also by known hospital patient movements after the outbreak onset. Furthermore, we were able to identify potential source cases and places just prior of the outbreak start. Unfortunately, we were not able to show the role of healthcare workers in MRSA transmission in a hospital setting due to limitations in data collection and network measure chosen (eigenvector centrality). Further research is required to confirm these study findings.