Yes / Objectives: This study presents isotopic information for incremental dentine collagen and bone bulk collagen from individuals from the Canary Islands (Tenerife and Gran Canaria) to explore dietary differences during childhood life. Materials and Methods: Eight individuals have been studied, which comprises 122 δ15N and δ13C incremental dentine measurements and eight bulk bone collagen analyses. A baseline of potentially consumed food sources has been developed for comparative purposes. A FRUITS model of probable contributions of each food source towards the diet of each individual has been developed. All samples but one belongs to the later period of indigenous occupation of the archipelago. Results: The dentine collagen data are presented in correlated δ13C and δ15N plots per individual, showing the isotopic changes throughout time. δ15N values for each individual tend to be variable whereas δ13C data are generally more stable with a range of +9.1 to +14‰ for δ15N and –17.4 to –20.8‰ for δ13C. Conclusion: The isotopic analysis allows for the reconstruction of 8 dietary profiles, which allow us to estimate the different dietary protein sources. The FRUITS model shows different percentages of the primary food sources for each individual. Where both δ13C and δ15N are elevated, this could be indicative of a higher marine contribution to the diet. There appear to be two main dietary profiles identifiable in the dataset and these may be related to changes in status or place of residence. Short-term variations in δ13C and δ15N and opposing co-variance of isotopic values can be indicative of nutritional stress, although metabolic changes during growth are also considered. / PhD grant from Universidad de La Laguna—Fundación La Caixa (“Contratos Predoctorales para la formación de doctores ULL-2015”), MINECO (Ministerio de Economía y Competitividad) project “Guanches y europeos en Las Cañadas del Teide, Ocupación, Producción y Comunicación” (HAR2015-68323-P), “Erasmus+ Scholarship” given by the University of La Laguna, ERC Starting Grant project IsoCAN (grant 851733, European Comission) and the projects RTI2018-101923-J-I00 and RYC2019-028346 (Ministerio de Ciencia e Innovación)

Industrial control systems (ICS) play an extremely important role in the world around us. They have helped in reducing human effort and contributed to automation of processes in oil refining, power generation, food and beverage and production lines. With advancement in technology, embedded platforms have emerged as ideal platforms for implementation of such ICSes. Traditional approaches in ICS design involve switching from a model or modeling environment directly to a real-world implementation. Errors have the potential to go unnoticed in the modeling environment and have a tendency to affect real control systems. Current models for error identification are complex and affect the design process of ICS appreciably. This thesis adds an additional layer to ICS design: an Interface Abstraction Process (IAP). IAP helps in incremental migration from a modeling environment to a real physical environment by supporting intermediate design versions. Implementation of the IAP is simple and independent of control system complexity. Early error identification is possible since intermediate versions are supported. Existing control system designs can be modified minimally to facilitate the addition of an extra layer. The overhead of adding the IAP is measured and analysed. With early validation, actual behavior of the ICS in the real physical setting matches the expected behavior in the modeling environment. This approach to ICS design adds a significant amount of latency to existing ICSes without affecting the design process significantly. Since the IAP helps in early design validation, it can be removed before deployment in the real-world. / Master of Science

This thesis presents a novel strategy to address the challenge of "catastrophic forgetting" in deep continual-learning systems. The term refers to severe performance degradation for older tasks, as a system learns new tasks that are presented sequentially. Most previous techniques have emphasized preservation of existing knowledge while learning new tasks, in some cases advocating a memory buffer that grows in proportion to the number of tasks. However, we offer another perspective, which is that mitigating local-task fitness during learning is as important as attempting to preserve existing knowledge. We posit the existence of a consistent, unlabelled world environment that the system uses as an easily-accessible reference to avoid favoring spurious properties over more generalizable ones. Based on this assumption, we have developed a novel method called Learning with Reference (LwR), which delivers substantial performance gains relative to its state-of-the-art counterparts. The approach does not involve a growing memory buffer, and therefore promotes better performance at scale. We present extensive empirical evaluation on real-world datasets. / Master of Science / Rome was not built in a day, and in nature knowledge is acquired and consolidated gradually over time. Evolution has taught biological systems how to address emerging challenges by building on past experience, adapting quickly while retaining known skills. Modern artificial intelligence systems also seek to amortize the learning process over time. Specifically, one large learning task can be divided into many smaller non-overlapping tasks. For example, a classification task of two classes, tiger and horse, is divided into two tasks, where the classifier only sees and learns from tiger data in the first task and horse data in the second task. The systems are expected to sequentially acquire knowledge from these smaller tasks. Such learning strategy is known as continual learning and provides three meaningful benefits: higher resource efficiency, a progressively better knowledge base, and strong adaptability. In this thesis, we investigate the class-incremental learning problem, a subset of continual learning, which refers to learning a classification model from a sequence of tasks. Different from transfer learning, which targets better performance in new domains, continual learning emphasizes the knowledge preservation of both old and new tasks. In deep neural networks, one challenge against the preservation is "catastrophic forgetting", which refers to severe performance degradation for older tasks, as a system learns new ones that are presented sequentially. An intuitive explanation is that old task data is missing in the new tasks under continual learning setting and the model is optimized toward new tasks without concerning the old ones. To overcome this, most previous techniques have emphasized the preservation of existing knowledge while learning new tasks, in some cases advocating old-data replay with a memory buffer, which grows in proportion to the number of tasks. In this thesis, we offer another perspective, which is that mitigating local-task fitness during learning is as important as attempting to preserve existing knowledge. We notice that local task data always has strong biases because of its smaller size. Optimization on it leads the model to local optima, therefore losing a holistic view that is crucial for other tasks. To mitigate this, a reliable reference should be enforced across tasks and the model should consistently learn all new knowledge based on this. With this assumption, we have developed a novel method called Learning with Reference (LwR), which posits the existence of a consistent, unlabelled world environment that the system uses as an easily-accessible reference to avoid favoring spurious properties over more generalizable ones. Our extensive empirical experiments show that it significantly outperforms state-of-the-art counterparts in real-world datasets.

Yes / Despite a growing literature, the relationship between financing constraints (FC), intellectual property rights (IPR) protection and firm innovation remains unclear within the transitional country context. Drawing on endogenous growth theory and extending the Gorodnichenko and Schnitzer (2013) framework, we hypothesize that in addition to firm-specific factors, country-level variables manifested within FC hamper incremental innovation, albeit in varying degrees due to industry heterogeneity. Secondly, as opposed to previous studies that solely focus on FC affecting firm innovation, we propose that due to resource constraints, firms in transition economies tend to follow an imitational innovation strategy, and therefore, from this perspective, IPR protection can be crucial for firm-level innovation within those economies. Using data from the World Bank Enterprise Survey (WBES) consisting of information for about 21,960 firms from 27 Eastern European and Central Asian transition countries and employing a two-step probit model with endogenous regressors, we find that adverse effects of FC and IPR on firms' innovation activities are driven from within as well as between industries. Focusing on the differential impacts of FC and IPR protection across industries, we direct potential causal pathways from easing FC and optimal IPR protection to encourage firms' innovation. Based on the findings, while very strict IPR protection is detrimental to firms' product and process innovation in industries with limited resource and skill capabilities, it is nevertheless helpful for research and development (R&D) activities in industries characterised by strong R&D and IP capacities. Our results offer useful insights for policymakers to support incremental innovation as well as boost invention. IPR protection policies require to be customised to the industries and firms, since invariably tight or lax IPR enforcement can be discouraging to both incremental and radical innovation, causing all industries suffering from the same treatment.

Yes / Background: Carbon (δ13C) and nitrogen (δ15N) isotope ratios of collagen from teeth and bone are used to study human nutrition and health. As bones are constantly remodelling throughout life, isotopic values of bone collagen represent an average of several years. In contrast, human teeth do not remodel and their primary dentine contains only the isotopic data from the time of formation. In contrast to all other bones, human auditory ossicles also appear not to remodel. As they develop in utero and finish formation in the first 2 years of life, their collagen should also represent isotopic values of these two relatively short periods. Aim: By comparing δ13C and δ15N data from ossicles and incremental dentine, this study aims to investigate how two developmental periods of the ossicles, in utero and the first 2 years of life, reflect in collagen obtained from the ossicles. Subject and methods: Ossicle and tooth samples of 12 individuals aged 0.5 ± 0.4 years to 13 ± 1 years from the nineteenth century St. Peter’s burial ground in Blackburn were collected and processed to obtain bulk bone and incremental dentine collagen which was measured for δ13C and δ15N. Results: Averaged δ13C and δ15N of ossicles are lower when compared to every age group except after 3 years of age. Average offset between ossicles and dentine of different groups ranges from 0.4–0.9‰ for δ13C and from 0.3–0.9‰ for δ15N, with highest counterbalance at birth and after the first 5 months after birth. Conclusions: There appears to be a systematic offset between the dentine and ossicle data. It seems that the second phase of development does not influence the isotopic values of collagen significantly and the data we are obtaining from ossicles represents the in utero period. / Research grant from The Society for the Study of Human Biology.

Yes / There are few reliable osteological indicators to detect parity or infer puberty in skeletal remains. Nitrogen (δ15N) and stable carbon (δ13C) isotope ratios in human tissues can be affected by metabolically unbalanced states engendered by pregnancy or rapid growth, offering potential biomarkers. This pilot study explores the potential of incremental dentine-collagen isotope ratio analysis to identify puberty and gestation. Methodology: Incremental dentine δ15N and δ13C profiles were produced by analysing third molars extracted as part of dental treatment of 10 individuals living in Sudan. Demographic and anthropometric data at the time of tooth extraction was available. Medical histories were unknown. Results: Isotopic signatures potentially related to pubertal growth, with an average δ15N reduction of 0.78±0.29‰, are indicated. Six isotopic signals suggestive of pregnancy, with an average δ15N decrease of 0.48±0.22‰, are also observed. The timing, speed and amplitude of post-partum δ15N patterns seemingly infer infant feeding practices and maternal nutritional status. Conclusion: This pilot study highlights the potential of incremental dentine isotope analysis for the reconstruction of early reproductive histories in skeletal remains. However, controlled studies with larger human cohort are needed to validate these findings, establish isotopic signals linked to puberty and lactation, and improve chronology accuracy.

Incremental sheet forming (ISF) is a novel metal shaping technology that is economically viable for low-volume manufacturing, customisation and rapid-prototyping. It uses a small tool that is controlled by a computer-numerically controlled sequence and the path taken by this tool over the sheet defines the product geometry. Little is currently known about how to design the tool-path to minimise geometric errors in the formed part. The work here addresses this problem by developing a model based tool-path optimisation scheme for ISF. The key issue is how to generate an efficient model for ISF to use within a path optimisation routine, since current simulation methods are too slow. A proportion of this thesis is dedicated to evaluating the applicability of the rigid plastic assumption for this purpose. Three numerical models have been produced: one based on small strain deformation, one based on limit analysis theory and another that approximates the sheet to a network of rods. All three models are formulated and solved as second-order cone programs (SOCP) and the limit analysis based model is the first demonstration of an upper-bound shell finite element (FE) problem solved as an SOCP. The models are significantly faster than commercially available FE software and simulations are compared with experimental and numerical data, from which it is shown the rigid plastic assumption is suitable for modelling deformation in ISF. The numerical models are still too slow for the path optimisation scheme, so a novel linearised model based on the concept of spatial impulse responses is also formulated and used in an optimal control based tool-path optimisation scheme for producing axisymmetric products with ISF. Off-line and on-line versions of the scheme are implemented on an ISF machine and it is shown that geometric errors are significantly reduced when using the proposed method. This work provides a new structured framework for tool-path design in ISF and it is also a novel use of feedback to compensate for geometrical errors in ISF.

A estampagem incremental é um processo de deformação incremental de chapa que pode ser classificado em três modos distintos: estampagem incremental sem matriz, estampagem incremental com matriz parcial e estampagem com matriz total. A deformação ocorre devido ao contato entre a ferramenta e a chapa, localmente e progressivamente, ao longo de uma trajetória pré-definida. A chapa permanece presa a suportes que fixam as suas extremidades de maneira a impedir os deslocamentos provocados pela passagem da ferramenta. A primeira parte deste trabalho compreende estudos em elementos finitos, usando o software LS-Dyna, sobre a trajetória das deformações na estampagem incremental para a fabricação de flanges em chapas previamente furadas de liga de alumínio AA1050. Foram realizados testes de contato e de linha que permitiram definir parâmetros da simulação. Os resultados para as deformações foram confrontados com os resultados obtidos experimentalmente por outros autores e permitiram verificar o comportamento das deformações nos flanges. As deformações crescem montonicamente até os valores máximos. Este resultado confirma a teoria de que na estampagem incremental de chapas a fratura ocorre sem o prévio aparecimento da estricção. A segunda parte deste trabalho trata do desenvolvimento de um suporte modular aplicável a estampagem incremental, tanto na sua variante sem matriz quanto com matriz, com o intuito de aumentar a flexibilidade do processo com relação à geometria da peça a ser deformada. Foi empregada a técnica de análise de valor objetivando alcançar a melhor funcionalidade do sistema. Como resultado foi obtido um suporte modular que atende os requisitos de rigidez, flexibilidade, modularidade, agilidade na troca da chapa a ser deformada e simplicidade construtiva. / The sheet incremental forming can be classified into three basic configurations, depending on the number of contact points between the sheet, tool and die (when present). Single incremental forming (absence of specific dies) and two point incremental forming (presence of a partial or full die) can be distinguished. The process makes use of a simple, low-cost, hemispherical tipped tool to precisely and progressively shape a blank into a metal sheet along a predefined tool path. The sheet is fixed onto the rig to prevent lateral displacements caused by tool movement. The first part of this work includes studies about finite elements, using the LS-Dyna software, applied to deformation path in hole-flanging produced by incremental sheet forming. The investigation was carried out using aluminum alloy AA1050. Contact and line numerical results support simulation parameters choices. The results from the strains were compared with empirical ones obtained by other authors. The results showed the deformation mechanics behavior of the hole-flanging. The strain paths of hole-flanging produced by incremental sheet forming grew linearly and monotonically from the origin to the maximum achievable strains. This result confirms the absence of local necking along plane strain directions. The second part of this work deals with the development of a modular rig that can be applied to dieless, partial or total die single point incremental forming in order to increase the flexibility of the process concerning the geometry of the part to be deformed. The result was a modular rig with constructive simplicity and requirements of stiffness, flexibility and modularity when exchanging sheets.

Este trabalho apresenta um modelo inovador de coletor solar plano, com placa absorvedora fabricada por Estampagem Incremental e Soldagem FSW (Friction Stir Welding). Esta placa absorvedora é constituída de duas chapas de alumínio AA1200-H14 com espessura de 1 mm, estampadas e soldadas em simetria, a fim de que na união das chapas sejam deixados canais para passagem de água. Neste estudo foi realizada a caracterização do material por Ensaio de Tração e Ensaio Nakajima. Para determinação dos parâmetros de Estampagem Incremental foram realizados 16 experimentos com ferramenta de diâmetro df = 9,5 mm, variando a rotação de N = 50 rpm a 800 rpm e o incremento vertical de Δz = 2 mm a 0,2 mm, mantendo o avanço em = 250 mm/min. E também foram realizados 3 experimentos com ferramenta df = 22 mm, variando o incremento vertical de Δz = 2 mm a 0,5 mm, mantendo a rotação em N = 50 rpm e o avanço em = 250 mm/min. Para determinação dos parâmetros de Soldagem FSW foram realizados 4 experimentos com ferramenta de ombro de diâmetro 8 mm e pino roscado M3x0,5, mantendo a rotação em N = 1500 rpm e variando o avanço entre = 100 mm/min a 400 mm/min. Em seguida foi fabricado um protótipo de placa absorvedora de coletor solar com área de 0,12 m². Nos experimentos realizados foi constatado que é possível obter maiores deformações na Estampagem Incremental do que na Estampagem Convencional e que as deformações são mais elevadas quando se utiliza menores diâmetros, maiores rotações e menores incrementos verticais da ferramenta. Na Soldagem FSW não foi constatada influência na qualidade do cordão de solda em relação à variação do avanço da ferramenta. Neste estudo também se verificou que é possível fabricar protótipos de placas absorvedoras de coletores solares pelos processos de Estampagem Incremental e Soldagem FSW. No entanto, para coletores em tamanho comercial, novos estudos são necessários para melhorar a forma de fixação das chapas durante a Soldagem FSW. / This work presents an innovative model of flat plate solar collector, with absorber plate manufactured using Incremental Sheet Forming (ISF) and Friction Stir Welding (FSW). This absorber plate consists of two AA1200-H14 aluminum sheets with a thickness of 1 mm, stamped and welded in symmetry, in order to leave channels for the passage of water. In this study the characterization of the material by Nakajima Test and Traction Test was performed. In order to determine the parameters of ISF, 16 experiments were performed with a tool of diameter df = 9.5 mm, varying the rotation speed of N = 50 rpm at 800 rpm and the step down of Δz = 2 mm to 0.2 mm, maintaining the feed rate at = 250 mm/min. Also, 3 experiments with tool df = 22 mm were performed, varying the step down of Δz = 2 mm to 0.5 mm, maintaining the rotation speed at N = 50 rpm and the feed rate at = 250 mm/min. For determination of FSW parameters, 4 experiments with 8 mm diameter shoulder tool and M3x0.5 pin were performed, maintaining the rotation speed at N = 1500 rpm and varying the feed rate from = 100 mm/min to 400 mm/min. A prototype solar collector absorber plate with a 0.12 m² area was then manufactured. In the experiments carried out, it was found that it is possible to obtain greater deformations in the ISF than in the Conventional Stamping and that the deformations are higher when using smaller diameters, higher rotations and smaller step downs of the tool. In FSW, no influence was observed in the quality of the weld bead in relation to the variation of the tool feed rate. In this study it was also verified that it is possible to manufacture prototypes of solar collector absorber plates by the processes of ISF and FSW. However, for commercial size collectors, further studies are needed to improve the way the plates are fixed during FSW.

Aprender conceitos provenientes de fluxos de dados é uma tarefa significamente diferente do aprendizado tradicional em lote. No aprendizado em lote, existe uma premissa implicita que os conceitos a serem aprendidos são estáticos e não evoluem significamente com o tempo. Por outro lado, em fluxos de dados os conceitos a serem aprendidos podem evoluir ao longo do tempo. Esta evolução é chamada de mudança de conceito, e torna a criação de um conjunto fixo de treinamento inaplicável neste cenário. O aprendizado incremental é uma abordagem promissora para trabalhar com fluxos de dados. Contudo, na presença de mudanças de conceito, conceitos desatualizados podem causar erros na classificação de eventos. Apesar de alguns métodos incrementais baseados no modelo de misturas gaussianas terem sido propostos na literatura, nota-se que tais algoritmos não possuem uma política explicita de descarte de conceitos obsoletos. Nesse trabalho um novo algoritmo incremental para fluxos de dados com mudanças de conceito baseado no modelo de misturas gaussianas é proposto. O método proposto é comparado com vários algoritmos amplamente utilizados na literatura, e os resultados mostram que o algoritmo proposto é competitivo com os demais em vários cenários, superando-os em alguns casos. / Learning concepts from data streams differs significantly from traditional batch learning. In batch learning there is an implicit assumption that the concept to be learned is static and does not evolve significantly over time. On the other hand, in data stream learning the concepts to be learned may evolve over time. This evolution is called concept drift, and makes the creation of a fixed training set be no longer applicable. Incremental learning paradigm is a promising approach for learning in a data stream setting. However, in the presence of concept drifts, out dated concepts can cause misclassifications. Several incremental Gaussian mixture models methods have been proposed in the literature, but these algorithms lack an explicit policy to discard outdated concepts. In this work, a new incremental algorithm for data stream with concept drifts based on Gaussian Mixture Models is proposed. The proposed methodis compared to various algorithms widely used in the literature, and the results show that it is competitive with them invarious scenarios, overcoming them in some cases.