As the Next-Generation (NextG, 5G and beyond) wireless network supports a wider range of services, optimization of resource allocation plays a crucial role in ensuring efficient use of the (limited) available network resources. Note that resource allocation may require knowledge of network parameters (e.g., channel state information and available power level) for package schedule. However, wireless networks operate in an uncertain environment where, in many practical scenarios, these parameters are unknown before decisions are made. In the absence of network parameters, a network controller, who performs resource allocation, may have to make decisions (aimed at optimizing network performance and satisfying users' QoS requirements) while emph{learning}. To that end, this dissertation studies two novel online learning problems that are motivated by autonomous resource management in NextG. Key contributions of the dissertation are two-fold. First, we study reward maximization under uncertainty with fairness constraints, which is motivated by wireless scheduling with Quality of Service constraints (e.g., minimum delivery ratio requirement) under uncertainty. We formulate a framework of combinatorial bandits with fairness constraints and develop a fair learning algorithm that successfully addresses the tradeoff between reward maximization and fairness constraints. This framework can also be applied to several other real-world applications, such as online advertising and crowdsourcing. Second, we consider global reward maximization under uncertainty with distributed biased feedback, which is motivated by the problem of cellular network configuration for optimizing network-level performance (e.g., average user-perceived Quality of Experience). We study both the linear-parameterized and non-parametric global reward functions, which are modeled as distributed linear bandits and kernelized bandits, respectively. For each model, we propose a learning algorithmic framework that can be integrated with different differential privacy models. We show that the proposed algorithms can achieve a near-optimal regret in a communication-efficient manner while protecting users' data privacy ``for free''. Our findings reveal that our developed algorithms outperform the state-of-the-art solutions in terms of the tradeoff among the regret, communication efficiency, and computation complexity. In addition, our proposed models and online learning algorithms can also be applied to several other real-world applications, e.g., dynamic pricing and public policy making, which may be of independent interest to a broader research community. / Doctor of Philosophy / As the Next-Generation (NextG) wireless network supports a wider range of services, optimization of resource allocation plays a crucial role in ensuring efficient use of the (limited) available network resources. Note that resource allocation may require knowledge of network parameters (e.g., channel state information and available power level) for package schedule. However, wireless networks operate in an uncertain environment where, in many practical scenarios, these parameters are unknown before decisions are made. In the absence of network parameters, a network controller, who performs resource allocation, may have to make decisions (aimed at optimizing network performance and satisfying users' QoS requirements) while emph{learning}. To that end, this dissertation studies two novel online learning problems that are motivated by resource allocation in the presence uncertainty in NextG. Key contributions of the dissertation are two-fold. First, we study reward maximization under uncertainty with fairness constraints, which is motivated by wireless scheduling with Quality of Service constraints (e.g., minimum delivery ratio requirement) under uncertainty. We formulate a framework of combinatorial bandits with fairness constraints and develop a fair learning algorithm that successfully addresses the tradeoff between reward maximization and fairness constraints. This framework can also be applied to several other real-world applications, such as online advertising and crowdsourcing. Second, we consider global reward maximization under uncertainty with distributed biased feedback, which is motivated by the problem of cellular network configuration for optimizing network-level performance (e.g., average user-perceived Quality of Experience). We consider both the linear-parameterized and non-parametric (unknown) global reward functions, which are modeled as distributed linear bandits and kernelized bandits, respectively. For each model, we propose a learning algorithmic framework that integrate different privacy models according to different privacy requirements or different scenarios. We show that the proposed algorithms can learn the unknown functions in a communication-efficient manner while protecting users' data privacy ``for free''. Our findings reveal that our developed algorithms outperform the state-of-the-art solutions in terms of the tradeoff among the regret, communication efficiency, and computation complexity. In addition, our proposed models and online learning algorithms can also be applied to several other real-world applications, e.g., dynamic pricing and public policy making, which may be of independent interest to a broader research community.

Modern mobile communication systems will provide enhanced high-speed data, multimedia, and voice services to mobile users. The integration of such heterogeneous traffic types implies that the network must provide differentiated Quality of Service (QoS). Beam forming techniques have been proposed to increase the spectral efficiency of the wireless channel. Using beamforming in the network will lead to intra-cell handoffs within the cell due to user mobility. In a commercially deployed next generation cellular system, it is likely that beam forming and QoS guarantees to the users will co-exist. In this work we propose a resource allocation and management scheme tailored for a network that employs smart antennas in support of a heterogeneous user mix. Resource management in a wireless system should take care of channel impairments and non-ideal antenna patterns. Mobile users moving from one beam to the other give rise to resource reallocation issues. Depending on the scatterer distribution in the cell, the Angle of Arrival (AoA) of the users will also change, affecting the interference pattern in the cell. In a system with data and multimedia users, some of the users are likely to be elastic in their demands for bandwidth. In this work, we propose a resource allocation and management scheme tailored for systems with smart antennas having heterogeneous users. The algorithm works by comparing the received power in the beams. Elasticity of user requirement for data services is exploited to provide adaptive QoS, thereby reducing the call dropping probability due to user mobility. Simulation results showing the channel and Multiple Access Interference (MAI) effects on system performance are presented. The effect of channel coding to provide Bit Error Rate (BER) guarantees is studied. We also show the throughput advantage obtained using the resource management algorithms. It is also seen that the throughput of the system increases for a user population having a higher elasticity. A modified resource allocation algorithm to reduce the blocking probability of the calls is presented and performance verified using simulation. The probability of call dropping in an unmanaged system due to user mobility is shown. Our studies show that using managed system the call drop probability can be minimized. / Master of Science

Tesis por compendio / During the last 50 years, there has been an overall improvement of efficiency in all domestic species thanks to an improvement in productivity traits. However, during the last years, disorders associated to a loss of functionality of animals (such as poor fertility, health and longevity of reproductive females) with high productivity standards have been observed. In this scenario, interactions between genetics and nutrition are becoming relevant. The main hypothesis of the present thesis is that nutritional strategies adapted to the genetic type could help to modulate productivity and functionality of females in the long term. To evaluate this hypothesis, 203 rabbit females were used which gave birth a total of 758 litters. These females belonged to 3 genetic types that differed greatly on their breeding goals: H, maternal line characterised by hyper-prolificacy; LP, maternal line characterised by functional hyper-longevity; R, paternal line characterised by growth rate. Females were fed during 5 consecutive reproductive cycles with 2 iso-energetic and iso-protein diets differing in energy source: Animal fat enhancing milk yield (84 g of ether extract per kg of DM and 105 g of starch per kg of DM); Cereal starch promoting body reserves recovery (21 g of ether extract per kg of DM and 237 g of starch per kg of DM). This experiment generated a set of longitudinal data with particular structures of (co)variances among data which had to be taken into account. Several models differing in the way data structure is modelled were evaluated. We found that none of the tested models was the best in terms of goodness of fit to the data for all traits, but it seems that models in which (co)variance structure was modelled in blocks of (co)variances for each reproductive cycle could be the most recommendable because they presented an overall suitable statistic performance (in terms of parsimony and fitting) and biological interpretation. These models provided useful information to understand acquisition and allocation of individual animals. In this sense, it seems that we can modulate allocation between energy addressed to milk or to body reserves by changing energy source of diet, but altering normal homeostasis of animals. Respect to the strategies of genetic types, results of the present experiment seem to indicate that R females were characterized by greater adult weight and by a high dependence on the body reserves to cope with the reproductive requirements of the current reproductive cycle. They gave birth little but heavier kits, although it seems they could be more immature. When R females were fed with a diet promoting milk yield, they invested more on the current litter, whereas when fed with a diet promoting body reserves recovery, it seems that they invested more in recovering for future reproduction. On the contrary, females from the maternal lines were smaller and had numerous but lighter kits, but each genetic type used different strategies. H females were also high dependent on body reserves, but storing body reserves during lactation to cope with future reproduction. This strategy makes them more sensible to the energy source of the diet, triggering problems to ensure future reproduction when fed with a diet promoting body reserves recovery (low conception rate or higher mortality of females). LP females were characterised by an acquisition capacity better fitted to changing requirements, safeguarding body reserves. This strategy seems to be more generalist, allowing them to ensure high performance of the current litter without neglecting future reproduction and with less sensitivity to the diet. Therefore, to properly balance between productivity and functionality, results reported in the presented thesis suggest that apart from requirements, the way each genetic type acquire and allocate energy over its life trajectory should be considered when formulating diets for reproductive rabbit females. / En los últimos 50 años, ha habido mejora general de la productividad y en consecuencia de la eficiencia en todas las especies domésticas. Sin embargo, en los últimos an¿os se han observado trastornos asociados a una pérdida de funcionalidad en animales con altos estándares de productividad (como baja fertilidad, salud o longevidad de las hembras reproductivas). Las interacciones entre la genética y la nutricio'n son cada vez ma's relevantes. La principal hipótesis de la presente tesis es que estrategias nutricionales adaptadas al tipo genético podrían ayudar a modular entre productividad y funcionalidad de las hembras. Para evaluar esta hipótesis, se utilizaron 203 conejas que dieron lugar a un total de 758 lactaciones. Estas conejas pertenecían a 3 tipos genéticos que diferían ampliamente en sus objetivos de selección: H, línea materna caracterizada por hiper-prolificidad; LP, línea materna caracterizada por hiper-longevidad funcional; R, línea paterna caracterizada por la velocidad de crecimiento. Las conejas fueron alimentadas durante 5 ciclos reproductivos consecutivos con 2 dietas iso-energéticas e iso-proteicas que diferían en la fuente de energía: Grasa animal para fomentar la producción de la leche (84 g de extracto etéreo por kg MS y 105 g de almidón por kg MS); Almidón de cereal para promover la recuperación de las reservas del corporales (21 g de extracto etéreo por kg MS y 237 g de almidón por kg MS). El experimento genero' una base de datos con datos longitudinales con unas estructuras particulares de (co)varianzas entre los datos y que debían tenerse en cuenta. En este sentido, se evaluaron varios modelos que diferían en la forma en que se modelan la estructura de datos. Se encontró que ninguno de los modelos probados fue el mejor en términos de bondad de ajuste para todos los caracteres. Sin embargo, parece que los modelos en los que la estructura (co) varianza se modelo' en bloques de (co)varianzas para cada ciclo reproductivo podrían ser los ma's recomendables. Respecto al efecto de la fuente de energía de la dieta, parece que podemos modular la asignación entre la energía dirigida a la producción de leche o la ganancia de reservas corporales cambiando la fuente de energía de la dieta, pero alterando la homeostasis normal de los animales. Respecto a las estrategias de los tipos genéticos, los resultados del presente experimento parecen indicar que las hembras R se caracterizaron por un mayor peso adulto y por una alta dependencia de las reservas corporales para hacer frente a las necesidades reproductivas del ciclo reproductivo actual. Estas hembras parieron menos gazapos, pero ma's pesados. Cuando las hembras R fueron alimentadas con una dieta que promovía la producción de la leche, invirtieron ma's en la camada actual, mientras que cuando se les alimento' con una dieta que promovía la recuperación de las reservas corporales, invirtieron ma's en la recuperación para su futura reproducción. Por el contrario, las hembras de las líneas maternas fueron ma's pequen¿as y tuvieron ma's gazapos, pero ma's ligeros. No obstante, cada tipo genético utilizo' una estrategia diferente. Las hembras H también fueron muy dependientes de las reservas corporales, pero almacenando reservas corporales durante la lactación para hacer frente a la futura reproducción. Esta estrategia les hace sensibles a la fuente de energía de la dieta, desencadenando problemas para asegurar la reproducción futura cuando fueron alimentadas con una dieta que promueve la recuperación de las reservas corporales. Las hembras LP se caracterizaron por una capacidad de adquisición mejor adaptada a las necesidades cambiantes. Esta estrategia parece ser ma's generalista, lo que les permite asegurar un alto rendimiento de la camada actual sin descuidar la reproducción futura, salvaguardando las reservas corporales, y con menos sensibilidad a la dieta. / En els últims 50 anys, ha aportat una millora general de la productivitat i, en conseqüència, de l'eficiència en totes les espècies domèstiques. Al llarg dels últims anys s'han observat trastorns associats a una pèrdua de funcionalitat en animals amb estàndards elevats de productivitat (com baixa fertilitat, salut o longevitat de les femelles reproductives). Les interaccions entre la genètica i la nutrició són cada vegada més rellevants. La hipòtesi principal de la tesi proposa que les estratègies nutricionals adaptades al tipus genètic permeten modular la productivitat i funcionalitat de les femelles. Per avaluar esta hipòtesi es van utilitzar 203 conilles que proporcionaren un total de 758 lactacions. Estes conilles pertanyien a tres tipus genètics que diferien àmpliament en els seus objectius de selecció: H, línia materna caracteritzada per hiperprolificitat; LP, línia materna caracteritzada per hiperlongevitat funcional; R, línia paterna caracteritzada per velocitat de creixement. A les conilles se'ls va subministrar 2 dietes isoenergètiques i isoproteiques durant cinc cicles reproductius consecutius, els quals diferien en la font d'energia: Greix animal per fomentar la producció de llet (84 g d'extracte eteri per kg MS i 105 g de midó per kg MS); Midó de cereal per promoure la recuperació de les reserves del corporals (21 g d'extracte eteri per kg MS i 237 g de midó per kg MS). L'experiment va generar una base de dades amb dades longitudinals amb unes estructures particulars de covariàncies entre les dades que havien de tindre's en compte. Es van avaluar diversos models que diferien en la forma en què modelitzaven l'estructura de les dades. Es va trobar que cap dels models provats va ser millor en termes de bondat d'ajust per a tots els caràcters. No obstant això, pareix que els models en què l'estructura covariància es va modelar en blocs de covariàncies per a cada cicle reproductiu podrien ser els més recomanables. Estos models van proporcionar informació útil per a comprendre l'adquisició i assignació d'animals individuals. Respecte a la font d'energia de la dieta, pareix que podem modular l'assignació entre l'energia dirigida a la producció de llet o el guany de reserves corporals si canvia la font d'energia de la dieta, tanmateix l'homeòstasi normal dels animals es veu alterada. Des del punt de vista de les estratègies segons el tipus genètics, els resultats del present experiment pareixen indicar que les femelles R es van caracteritzar per tindre major pes adult i per una dependència elevada de les reserves corporals per a fer front a les necessitats reproductives del cicle reproductiu actual. Estes femelles van parir menys catxaps. Quan les femelles R van ser alimentades amb una dieta que promovia la producció de la llet, van invertir més en la ventrada actual, mentres que quan se'ls va alimentar amb una dieta que promovia la recuperació de les reserves corporals, van invertir més en la recuperació per al següent cicle reproductiu. Contràriament, les femelles de les línies maternes van ser més menudes i van tindre més catxaps, però més lleugers. No obstant això, cada tipus genètic va utilitzar una estratègia diferent. Les femelles H també van ser molt dependents de les reserves corporals, però emmagatzemaren reserves corporals durant la lactació per a fer front a pròxima reproducció. Esta estratègia les fa més sensibles a la font d'energia de la dieta, i conseqüentment, desencadena problemes per assegurar la reproducció propera quan es nodrixen amb una dieta que promou la recuperació de les reserves corporals (baixa taxa de concepció o major mortalitat de les femeelle). Les femelles LP es van caracteritzar per una capacitat d'adquisició de recursos millor adaptada a les necessitats canviants. Esta estratègia pareix més generalista, la qual cosa les permet assegurar un rendiment alt de la ventrada actual sense descuidar la reproducció fu / Arnau Bonachera, A. (2017). OPTIMIZATION OF RESOURCE ALLOCATION TO THE GENETIC TYPE IN REPRODUCTIVE RABBIT DOES [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/85980 / Compendio

付記する学位プログラム名: 京都大学卓越大学院プログラム「先端光・電子デバイス創成学」 / 京都大学 / 新制・課程博士 / 博士(情報学) / 甲第25445号 / 情博第883号 / 新制||情||148(附属図書館) / 京都大学大学院情報学研究科通信情報システム専攻 / (主査)教授 大木 英司, 教授 原田 博司, 教授 岡部 寿男 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Protection of critical national infrastructure has received significant attention in the past recent years. As a result, researchers have developed methods to preserve and maintain critical infrastructure systems and minimize their vulnerability to disasters and disruptions. However, these models are often customized to meet the characteristics and functionality requirements for a particular system, and are computationally intensive and require simplifying assumptions. In this study, we first develop a tractable and relatively comprehensive model for optimizing maintenance planning of generic network-structured systems. We considered both linear and nonlinear objective functions for our problems. We then reformulate the model in order to enhance its computational effectiveness for large scale complex problems. The proposed modeling framework inherently captures the network topography, the stochastic nature of disruptions, and can be applied to network-structured systems for which performance is assessed based on network flow efficiency and mobility. A hypothetical small-sized network is used to illustrate the developed models, and the pro- posed models are also applied to analyze a larger scale real network in order to assess their relative computational effectiveness and robustness. We selected the Istanbul highway net- work for the latter purpose because of its critical location and also because it has been considered in previous studies, which enables us to compare the effectiveness of our models with an existing model in the literature. We designed several test cases (considering single and multiple treatment types, and linear and nonlinear objectives), and solved them on the NEOS server using different available software. The results demonstrate that our models are capable of attaining optimal solutions within a very short time. Furthermore, the linear model is shown to yield a good approximation to the nonlinear model (it determined solutions within 0.3% of optimality, on average). Moreover, in both cases (our hypothetical illustrative example and the Istanbul highway network), the optimal policies obtained appear to favor the selection of fewer links and to apply a higher quality treatment to them. / Master of Science

An algorithmic approach is developed for solving the minimax continuous resource allocation problem with multimodal cost functions. Unlike previous research in the same area which developed solutions for the same problem by imposing restrictions on the cost functions, such as the assumptions of monotoniciy or convexity, this approach is applicable to problems with multimodal functions with a finite number of local extrema. Another significant advantage demonstrated by this approach is that it provides all the optimal solutions to the problem; in contrast to previous algorithms which provided a single optimal solution. When a further level of optimization using a second objective function is desired, one needs the entire set of optimal solutions as provided by the procedures of this thesis. / Master of Science

This research was initiated to address a shortfall in some of the previous work reviewed with respect to capital budgeting and allocation. This shortfall has to do with the very general manner in which the public sector is addressed. The public sector poses problems which are unique to its environment and the simple extension of private sector tools to this domain is considered ineffective. The central aim of this research is to develop an initial methodology for capital allocation to projects in the public sector. In developing this methodology the fundamental differences between the public and private sectors are pointed out. The methodology devised utilizes knowledge from the fields of Engineering Economy, Capital Budgeting, and Multi-Attribute Decision Analysis. Basic principles of Engineering Economy and Capital Budgeting are reviewed. The Multi-Attribute Decision Model selected for implementation in this research methodology is the Analytic Hierarchy Process (AHP). The procedure followed by the AHP is presented and its benefits and limitations are discussed. The research addresses the quantitative and qualitative aspects of assessing projects in the public sector. A hypothetical example is presented to demonstrate the manner in which the methodology works. The thesis concludes with recommendations and proposes areas for further research. / Master of Science

This thesis is concerned with the development of heuristic algorithms for the popular Quadratic Assignment Problem (QAP) which finds a wide variety of applications in various fields. This discrete optimization problem, which seeks the placement of m facilities on m locations in order to minimize a quadratic interactive cost, is well known to be NP-hard and turns out to be computationally intractable for even moderately sized problems. Hence, problems involving more than 12-15 facilities usually need to be analysed by approximate solution procedures. The more successful heuristic procedures which exist for problem QAP are computationally intensive, some of these resulting from a premature termination of exact solution procedures. The motivation here is to develop a polynomial time heuristic which is effective with respect to the quality of solutions obtained, while at the same time not being computationally very expensive. The method proposed herein is flexible in that one can operate it to suitably trade solution quality against effort as desired, and is portable in that the modules used as building blocks can be employed in conjunction with other heuristics as well. Computational experience on test problems found in the literature is provided to evaluate the worth of this method. / M.S.

Background: Insufficient resources and dynamic infection control policies during the COVID-19 pandemic created a resource-strained environment which necessitated frontline Health Care Providers (HCPs) to make ethical decisions frequently. Many of these ethical decisions included allocating scarce resources to optimally prioritize patients, resources, and clinician time. The transition from usual patient-centred care to care centred around infection control mandates and rationing resources forced HCPs to balance competing demands while trying to uphold high standards of care. This research aimed to understand the resource allocation decisions HCPs had to navigate during the pandemic and the ethical considerations guiding them. Methods: Using a qualitative case study approach, we aimed to document the type of ethical decision, reasoning used, and the action frontline HCPs took during the pandemic. Twenty-five semi-structured interviews were conducted with multi-disciplinary HCPs employed in a single community Intensive Care Unit (ICU) in Ontario. Resource allocation decisions were extracted from the transcripts and were analyzed using conventional content analysis. Results: Resource allocation decisions within critical care practice were ubiquitous and diverse. The constraints imposed by the pandemic and multiple provincial and organizational policies formed the context that necessitated these decisions. HCPs drew upon a range of ethical theories, notably Utilitarianism and Virtue Ethics, while prioritizing HCP safety and patient well-being. Resulting actions included prioritizing clinical tasks, establishing boundaries, and adapting practice patterns. Although these situations commonly evoked stress and frustration amongst HCPs, some positive internal responses were also described, including feelings of self-efficacy, resourcefulness, and team cohesion. Conclusion: In conclusion, analysis of resource allocation-derived decision-making illuminated a variety of challenges that HCPs faced during the COVID-19 pandemic, driven by institutional policies and pragmatic limitations. Insights from this study underscore how these ethical decisions are an inherent part of clinical practice and have the potential to foster positive professional development amidst adversity. / Thesis / Master of Health Sciences (MSc) / During the COVID-19 pandemic, Health Care Providers (HCPs) faced difficult choices due to limited resources and evolving infection control rules. HCPs had to decide how to allocate scarce resources and balance competing demands while maintaining high standards of care. This research aimed to understand the types of resource allocation decisions HCPs made and the ethical considerations guiding them. Through twenty-five interviews with HCPs from an Intensive Care Unit in Ontario, we found that such decisions were both frequent and diverse. HCPs considered ethical theories like Utilitarianism and Virtue Ethics, prioritizing patient well-being and HCP safety. They often had to adapt their practices, establish boundaries, and determine priorities. Although these decisions caused stress and frustration, they also fostered team cohesion and personal growth. Overall, this research sheds light on some challenges HCPs faced during the COVID-19 pandemic and underscores the importance of preparing learners to navigate ethical decision-making in clinical practice.

As the leader of an audit team, the audit partner can have a significant impact on the outcomes of an audit engagement. One hitherto unexplored aspect of partner assignments to audits is cases where they handle clients in multiple audit offices. I examine factors associated with the assignment of audit partners to multiple offices (henceforth multi-office partners or MOPs) and the implications of such assignments for audit quality. I document that audit firms assign partners to multiple offices to match partner expertise to client needs and to manage resource constraints in audit offices. Specifically, partners specializing in the financial sector, who report more skills on LinkedIn profiles, and have more professional experience, are more likely to be MOPs. Audit offices with fewer partner resources are more likely to share partners among the network of offices to mitigate resource constraints. In the audit quality analyses, I find that MOPs are, on average, associated with an increased likelihood of restatements, suggesting the negative impact of audit office resource constraints dominates the positive influence of partner expertise on audit quality. The negative effect of MOPs on audit quality exists for both local and non-local clients and is concentrated in MOPs who (1) are not industry specialists, (2) face greater information friction as the distance between different audit offices increases, and (3) have limited knowledge-sharing opportunities from audit offices. / Business Administration/Accounting