Browsing by Title

Traditional flat classification methods (e.g., binary, multiclass, and multi-label classification) seek to associate each example with a single class label or a set of labels without any structural dependence among them. Although, there are some problems in which classes can be divided or grouped into subclasses or superclasses respectively. Such a scenario demands the application of methods prepared to deal with hierarchical classification. An algorithm for hierarchical classification uses the information related to structure present in the class hierarchy and then improves the predictive performance . The freedom to perform a more generic classification, but with higher reliability, gives the process a greater versatility. Several studies have shown that, in solving a hierarchical classification problem, flat models are mostly overcome by hierarchical ones, regardless of the approach – local (including its derivations) or global – chosen. This thesis aims to compare the most popular hierarchical classification methods (local and global) empirically, reporting their performance – measured using hierarchical evaluation indexes. To do so, we had to adapt the global hierarchical models to conduct single path predictions, starting from the root class and moving towards a leaf class within the hierarchical structure. Further, we applied hierarchical classification on data streams by detecting concept drift. We first study data streams, various types of concept drifts, and state-of-the-art concept drift detection methods. Then we implement Global-Model Hierarchical Classification Naive Bayes (GNB) with three concept drift detectors: (i) Kolmogorov-Smirnov test, (ii) Wilcoxon test, and (iii) Drift Detection Method (DDM). A fixed-size sliding window was used to estimate the performance of GNB online. Finally, we must highlight that this thesis contributes to the task of automatic insect recognition.

Text classification, also known as text categorization, is a task to classify documents into predefined sets. As the prosperity of the social networks, a large volume of unstructured text is generated exponentially. Social media text, due to its limited length, extreme imbalance, high dimensionality, and multi-label characteristic, needs special processing before being fed to machine learning classifiers. There are all kinds of statistics, machine learning, and natural language processing approaches to solve the problem, of which two trends of machine learning algorithms are the state of the art. One is the large-scale linear classification which deals with large sparse data, especially for short social media text; the other is the active deep learning techniques, which takes advantage of the word order. This thesis provided an end-to-end solution to deal with large-scale, multi-label and extremely imbalanced text data, compared both the active trends and discussed the effect of balance learning. The results show that deep learning does not necessarily work well in this context. Well-designed large linear classifiers can achieve the best scores. Also, when the data is large enough, the simpler classifiers may perform better.

This study presents some of the available methods for haplotype reconstruction and evaluates the accuracy and efficiency of three different software programs that utilize these methods. The analysis is performed on the QTLMAS XII common dataset, which is publicly available. The program LinkPHASE 5+, rule-based software, considers pedigree information (deduction and linkage) only. HiddenPHASE is a likelihood-based software, which takes into account molecular information (linkage disequilibrium). The DualPHASE software combines both of the above mentioned methods. We will see how usage of different available sources of information as well as the shape of the data affects the haplotype inference.

Air ions can play an important role in new particle formation (NPF) process and consequently influence the atmospheric aerosols, which affect climate and air quality as potential cloud condensation nuclei. However, the air ions and their role in NPF have not been comprehensively investigated yet, especially in polluted area. To explore the air ions in polluted environment, we compared the air ions at SORPES site, a suburban site in polluted eastern China, with those at SMEAR II, a well-studied boreal forest site in Finland, based on the air ion number size distribution (0.8-42 nm) measured with Neutral Cluster and Air Ion Spectrometer (NAIS) during 7 June 2019 to 31 August 2020. Air ions were size classified into three size ranges: cluster (0.8-2 nm), intermediate (2-7 nm), and large (7-20 nm). Median concentration of cluster ions at SORPES (217 cm−3) was about 6 times lower than that at SMEAR II (1268 cm−3) due to the high CS and pre-existing particle loading in polluted area, whereas the median large ion concentration at SORPES (197 cm−3) was about 3 times higher than that of SMEAR II (67 cm−3). Seasonal variations of ion concentration differed with ion sizes and ion polarity at two sites. High concentration of cluster ions was observed in the evening in the spring and autumn at SMEAR II, while the cluster ion concentration remained at a high level all day in the same seasons. The NPF events occurred more frequently at SORPES site (SMEAR II 16% ; SORPES: 39%), and the highest values of NPF frequency at both sites were in spring ((SMEAR II: spring: 43%; SORPES: spring: 56%). During the noon time on NPF event day, the concentration of intermediate ions were 8-14 times higher than same ours on non-event days, indicating that can be used as an indicator for NPF in SMEAR II and SORPES. The median formation rate of 1.5 nm at SMEAR II were higher then that at SORPES, while higher formation rate of 3 nm ions were observed at SORPES. At 3 nm, the formation rate of charged particles was only 11% and 1.6% of the total rate at SMEAR II and SORPES respectively, which supports the current view that neutral ways dominate the new particle process in continental boundary. However, higher ratio between charged and total formation rate of 3 nm particle at SMEAR II indicates ion-induced nucleation can have a bigger contribution to NPF in clear area in comparison to polluted area. Higher median GR of 3-7 nm (SMEAR II: 3.1 nm h−1; SORPES: 3.7 nm h−1) and 7-20 nm (SMEAR II: 5.5 nm h−1; SORPES: 6.9 nm h−1) ions at SORPES were found in comparison to SMEAR II, suggesting the higher availability of condensing vapors at SORPES. This study presented a comprehensive comparison of air ions in completely different environments, and highlighted the need for long-term ion measurements to improve the understanding of air ions and their role in NPF in polluted area like eastern China

Due to its exceptional thermal properties and irradiation resistance, tungsten is the material of choice for critical plasma-facing components in many leading thermonuclear fusion projects. Owing to the natural retention of hydrogen isotopes in materials such as tungsten, the safety of a fusion device depends heavily on the inventory of radioactive tritium in its plasma-facing components. The proposed methods of tritium removal typically include thermal treatment of massive metal structures for prolonged timescales. A novel way to either shorten the treatment times or lower the required temperatures is based performing the removal under an H-2 atmosphere, effectively exchanging the trapped tritium for non-radioactive protium. In this thesis, we employ molecular dynamics simulations to study the mechanism of hydrogen isotope exchange in vacancy, dislocation and grain boundary type defects in tungsten. By comparing the results to simulations of purely diffusion-based tritium removal methods, we establish that hydrogen isotope exchange indeed facilitates faster removal of tritium for all studied defect types at temperatures of 500 K and above. The fastest removal, when normalising based on the initial occupation of the defect, is shown to occur in vacancies and the slowest in grain boundaries. Through an atom level study of the mechanism, we are able to verify that tritium removal using isotope exchange depends on keeping the defect saturated with hydrogen. This study also works to show that molecular dynamics indeed is a valid tool for studying tritium removal and isotope exchange in general. Using small system sizes and spatially-parallelised simulation tools, we have managed to model isotope exchange for timescales extending from hundreds of nanoseconds up to several microseconds.

We introduce a new model for contingent convertibles. The write-down, or equity conversion, and default of the contingent convertible are modeled as states of conditional Markov process. Valuation formulae for different financial contracts, like CDS and different types of contingent convertibles, are derived. The Model can be thought of as an extension to reduced form models with an additional state. For practical applications, this model could be used for new type of contingent convertible derivatives in a similar fashion than reduced form models are used for credit derivatives.

In recent years, the concept of Metaverse has become a popular buzzword in the media and different communities. In 2021, the company behind Facebook rebranded itself into Meta Platforms, inc. in order to match their new vision of developing the Metaverse. The Metaverse is becoming reality as intersecting technologies, including head-mounted virtual reality displays (HMDs) and non-fungible tokens (NFTs), have been developed. Different communities, such as media, researchers, consumers and companies have different perspectives on the Metaverse and its opportunities and problems. Metaverse technology has been researched thoroughly, while little to none research has been done on gray literature, i.e. non-scientific sources, to gain insight on the ongoing hype. The conducted research analyzed 44 sources in total, ranging from news articles to videos and forum discussions. The results show that people are seeing opportunities in Metaverse entrepreneurship in the changing career landscape. However, the visions of Meta Platforms, inc. also receive a fair amount of critique in the analyzed articles and threads. The results suggest that most of the consumers are only interested in a smaller subset of features than what is being marketed. The conducted research gives insight on how different sources are seeing the Metaverse and can therefore be used as a starting point for more comprehensive gray literature studies on the Metaverse. While making innovations to the underlying technology is important, studying people’s viewpoints is a requirement for the academia to understand the phenomenon and for the industry to produce a compelling product.

The thesis is written on the catalytic activation of carbon dioxide. It includes a literature part and an experimental part. In the literature part, a review on metal (salen) complexes in relation to their electronic and geometric properties is presented. Salicylidene-aminates are included considering similarity to the salens. Also included from literature is a selective review focusing on the mechanistic aspects in the carboxylation of epoxides by metal (salen) complexes. Some applications of iron (salen) complexes as catalyst are mentioned. In the experimental part, the bis(phenoxyiminato) chlorido iron(III)complexes are synthesized, characterized and applied on carbon dioxide/epoxide coupling reactions. Characterization is done by UV-vis, infra-red, nuclear magnetic resonance and electron impact mass spectroscopy and elemental analysis for C, H, and N. Thermogravimetric analysis for the complexes, DFT calculation for the most active species (L11)2Fe(III)Cl and X-ray for (L6)2Fe(III)Cl are also presented. X-ray crystallography reveals the space group of (L6)2Fe(III)Cl to be orthorhombic, Pbcn; a = 29.0038(14) Å, b = 8.6123(8) Å, c = 10.7843(9) Å; α = β = γ = 90 o. The ML2Cl complexes are observed to have M-O and M-N bonds involving the phenolic oxygen and azomethine nitrogen. Correlation study between spin state and the Fe-N bond length indicates a high-spin state for Fe(III) nucleus. The geometry around the metal nucleus is distorted square pyramidal. Reaction conditions for catalytic activity were fine-tuned envisaging the exclusive production of cyclic carbonates. Propylene and styrene oxides show high reactivity. The ketiminato complexes show better activity over the aldiminato complexes. Optimal result is obtained in dimethyl formamide at a temperature of 145 oC and carbon dioxide pressure of 10 bars in the presence of tetrabutylphosphonium bromide as co-catalyst. A TOF of 572/h is observed for propylene oxide. Three reaction mechanisms are proposed. Comparatively the Co(III) analogues are more active, and iodide as a halogen ligand produces a more active complex than chloride. Improving nucleophilicity of Fe(III), elimination of intramolecular H-bond and improving on solubility could yield a more active complex. Iron is a cheap and environmentally benign metal. The use of iron complexes is an attractive alternative to other transition metals which are expensive and/or toxic. The complexes are robust and show high thermal stability. Surprisingly, oligomers of styrene carbonate were noticed at the reaction temperature and pressure known to favor exclusive production of cyclic carbonates. These observations suggest the complexes for a promising study and application in future research for copolymerization. Such copolymers may have useful charateristics for diverse applications.

A Bayesian Network (BN) is a graphical model applying probability and Bayesian rule for its inference. BN consists of structure, that is a directed acyclic graph (DAG), and parameters. The structure can be obtained by learning from data. Finding an optimal BN structure is an NP-Hard problem. If an ordering is given, then the problem becomes simpler. Ordering means the order of variables (nodes) for building the structure. One of structure learning algorithms that uses variable ordering as the input is K2 algorithm. The ordering determines the quality of resulted network. In this work, we apply Cuckoo Search (CS) algorithm to find a good node ordering. Each node ordering is evaluated by K2 algorithm. Cuckoo Search is a nature-inspired metaheuristic algorithm that mimics the aggressive breeding behavior of Cuckoo birds with several simplifications. It has outperformed Genetic Algorithms and Particle Swarm Optimization algorithm in finding an optimal solution for continuous problems, e.g., functions of Michalewicz, Rosenbrock, Schwefel, Ackley, Rastrigin, and Griewank. We conducted experiments on 35 datasets to compare the performances of Cuckoo Search to GOBNILP that is a Bayesian network learning algorithm based on integer linear programming and it is well known to be used as benchmark. We compared the quality of obtained structures and the running times. In general, CS can find good networks although all the obtained networks are not the best. However, it sometimes finds only low-scoring networks, and the running times of CS are not always very fast. The results mostly show that GOBNILP is consistently faster and can find networks of better quality than CS. Based on the experiment results, we conclude that the approach is not able to guarantee obtaining an optimal Bayesian network structure. Other heuristic search algorithms are potentially better to be used for learning Bayesian network structures that we have not compared to our works, for example the ordering-search algorithm by Teyssier and Koller [41] that combines greedy local hill-climbing with random restarts, a tabu list, caching computations, and a heuristic pruning procedure.

Cancer consists of heterogeneous cell populations that repeatedly undergo natural selection. These cell populations contest with each other for space and nutrients and try to generate phenotypes that maximize their ecological fitness. For achieving this, they evolve evolutionarily stable strategies. When an oncologist starts to treat cancer, another game emerges. While affected by the cellular evolution processes, modeling of this game owes to the results of the classical game theory. This thesis investigates the theoretical foundations of adaptive cancer treatment. It draws from two game theoretical approaches, evolutionary game theory and Stackelberg leader-follower game. The underlying hypothesis of adaptive regimen is that the patient's cancer burden can be administered by leveraging the resource competition between treatment-sensitive and treatment-resistant cells. The intercellular competition is mathematically modelled as an evolutionary game using the G function approach. The properties of the evolutionary stability, such as ESS, the ESS maximum principle, and convergence stability, that are relevant to tumorigenesis and intra-tumoral dynamics, are elaborated. To mitigate the patient's cancer burden, it is necessary to find an optimal modulation and frequency of treatment doses. The Stackelberg leader-follower game, adopted from the economic studies of duopoly, provides a promising framework to model the interplay between a rationally playing oncologist as a leader and the evolutionary evolving tumor as a follower. The two game types applied simultaneously to cancer therapy strategisizing can nourish each other and improve the planning of adaptive regimen. Hence, the characteristics of the Stackelberg game are mathematically studied and a preliminary dose-optimization function is presented. The applicability of the combination of the two games in the planning of cancer therapy strategies is tested with a theoretical case. The results are critically discussed from three perspectives: the biological veracity of the eco-evolutionary model, the applicability of the Stackelberg game, and the clinical relevance of the combination. The current limitations of the model are considered to invite further research on the subject.