This thesis is about building a model that creates specific types of networks. A network, in this case, is simply a collection of elements (like people) and the connections between them (like friendships or communication). Mathematically, we call this a graph. There already exists a method called the configuration model that randomly builds such graphs while following certain rules—such as how many connections each person should have. When we use this model to make very large graphs, something interesting happens: a large, connected group of ‘people’ almost always forms. We call this the giant component, and it’s a bit like the main cluster in a social network where everyone is connected, directly or indirectly. For a given setup, the giant component has a predictable–typical– size. It’s extremely rare—-almost impossible–for it to be much smaller or larger than typically. But sometimes, we want to study what happens in those rare cases. This thesis explores how to build a new model that can generate these unusual graphs on purpose; graphs that still look natural, but where the giant component is either larger or smaller than the typical giant. This allows us to study networks in extreme scenarios that we normally can’t access.

Migraine is a prevalent, complex neurovascular disorder that mainly affects women. The exact pathophysiology of migraine is unclear, but research indicates that activation of the trigeminal nerve in the trigeminovascular system causes release of calcitonin gene-related peptide (CGRP), which triggers migraine attacks. Additionally, nitric oxide (NO) contributes to the pathophysiology of migraine headaches. Research suggests that migraine patients have an increased risk of cardiovascular disease. Over the past decades, non-invasive techniques, like Laser Doppler imaging (LDI) and Laser speckle contrast imaging (LSCI) have been developed for imaging tissue perfusion, which are valuable tools for investigating the underlying causes for this increased risk and facilitates the study of blood perfusion. The Erasmus Medical Center (Erasmus MC) used these techniques to perform measurements of the microvascular blood flow in the forearm as a measure of the microvasculature of women with and without migraine. This study, also known as the VASCULAR-study, focused on three regions of interest (ROIs), where NO was inhibited using iontophoresis with L-NMMA and neuropeptides were blocked using EMLA cream. The last ROI served as the control region. The measurements also consisted of three different phases: baseline, peak and plateau phase. Studies have shown that distinct biological mechanisms in the body can be linked to different frequency intervals and therefore, Fourier analysis (FA) and wavelet analysis (WA) were used to transform the VASCULAR-study data into the frequency domain. Two preliminary studies used FA to perform the transformation. This study primarily focused on WA. The research question is as follows: Does wavelet analysis (WA) of the VASCULAR-study data yield more insights than Fourier analysis (FA) into blood flow measurements in women, particularly in examining the role of nitric oxide (NO) and calcitonin gene-related peptide (CGRP) in the microvasculature among women with and without migraine? To address the research question, WA was conducted using the complex Morlet wavelet. Relative energy density was used as a quantitative metric to compare the group of women with migraine with the group of women without migraine. Relative energy density was also calculated for the results using FA. Statistical significance was assessed using p-values, where p-values below 0.05 were considered significant. The Mann-Whitney U-Test and the Wilcoxon Signed Rank Test were used to calculate the p-values. Significant differences between women without migraine and women with migraine were primarily found in respiratory and endothelial activity for both WA and FA. Women with migraine showed higher respiratory activity in regions where NO was inhibited, for both WA and FA. Although WA and FA revealed many similar results in the VASCULAR-study dataset, there were also some differences. These differences were mainly observed in endothelial activity, in the ROI where NO was inhibited. FA revealed significantly higher values in activity for women without migraine in both NO-independent and NO-dependent endothelial activity. Furthermore, using the time-frequency localization capability of the WA, it showed significantly higher activity in women with migraine between the peak and plateau phase.

In modern data analysis, high-dimensional datasets where the number of features far exceeds the number of observations are increasingly common. In such settings, identifying sparse informative features is a critical challenge. This thesis investigates the comparative performance of random forests and two penalized linear regression techniques, Ridge and Lasso regression, in scenarios with few informative features. Synthetic datasets were generated to simulate both linear and non-linear relationships between features and the target variable. The models were evaluated using mean squared error (MSE) to measure prediction accuracy and two custom metrics assessing the ability to identify informative features among the top-ranked variables. Additionally, computational efficiency was assessed.
In linear settings, Lasso consistently outperformed the other models in both prediction and feature selection, particularly at the lower informative ratios. Ridge regression demonstrated reasonable feature selection accuracy in linear settings but underperformed in prediction, likely due to a limited hyperparameter grid. In non-linear settings, random forests were most effective at identifying informative features, though the prediction performance of Lasso remained competitive and slightly outperformed random forests under high sparsity. Training time analysis further showed that the penalized linear models were substantially faster than randomforests. These findings underscore that no single model is universally optimal, as performance varies depending on datastructure, sparsity, and the modelling objective. They also demonstrate that improved feature selection does not necessarily guarantee better prediction in complex, high-dimensional settings. These results have practical implications for domains like genomics, where balancing accuracy, interpretability, and computational efficiency is critical.

In the last decade the graph burning model was developed to model the spread of information between people. Graph burning is a process that is done in rounds with the aim of spreading information to every connected person in a network. Every round one new source of information may be appointed and information spreads from people who have received it, to all of their connections, just like fire spreads. The burning number of a graph, denoted by 𝑏(𝐺), is the parameter that quantifies the speed of this spread of information. It has been conjectured that the burning number for a connected graph on 𝑛 vertices is at most ⌈√𝑛⌉. We prove the burning number conjecture for cat-constructs. Cat-constructs are trees obtained from a path graph 𝑃𝑛 by adding at most two vertices to subtrees of 𝑃𝑛. We show the burning sequence of a cat-construct may contain one fewer source than its burning number if the number of vertices for the cat-construct is more than the first square bigger than 𝑛. With this result

Hoewel het probleemoplossen in het wiskundecurriculum is opgenomen, hebben docenten vaak behoefte aan meer praktische ondersteuning bij het onderwijzen ervan. De recent in het Nederlands gepubliceerde didactiek ”Building Thinking Classrooms” van Peter Liljedahl met daarin veertien praktisch beschreven principes kan daarbij van betekenis zijn.

Het doel van dit onderzoek is om meer inzicht te verkrijgen in de effectiviteit van twee van de veertien principes ten aanzien van de beoogde leeropbrengst binnen de specifieke lespraktijk. Hierbij gaat het om het maken van aantekeningen met behulp van een ontworpen werkblad en het consolideren van een les door ontworpen lesactiviteiten waarin leerlingen zelfstandig reflecteren. De specifieke lespraktijk waarin dit onderzoek wordt uitgevoerd bestaat uit zes onafhankelijke lessen in ”Thinking Classrooms”. Deze lessen staan in het teken van het onderwijs van het oplossen van algemene wiskundige problemen. De bijbehorende leerdoelen zijn gerelateerd aan het ontwikkelen van metacognitieve vaardigheden en een geschikte wiskundige houding, wat volgens Alan Schoenfeld behoren tot twee van de vier essentiële bekwaamheden voor het wiskundig probleemoplossen. Om het doel van dit onderzoek te bereiken is de volgende hoofdvraag opgesteld.

Welke leerlingactiviteiten binnen de principes “note-taking” en het “consolideren” in een “Thinking Classroom”, ontwikkeld door Peter Liljedahl, dragen bij aan een verhoogde leeropbrengst met betrekking tot het oplossen van algemene wiskundige problemen door leerlingen een gemengde havo/vwo bovenbouwklas?

Dit onderzoek is uitgevoerd door middel van het cyclische ”Design-Based research”. Deze onderwijsbenadering richt zich richt op het ontwerpen, implementeren en evalueren van onderwijsinterventies in echte leeromgevingen. Naar aanleiding van de verschillende iteraties kunnen er patronen worden ontdekt in het leergedrag van de leerlingen. Zo blijkt er dat een individueel werkblad, na het gezamenlijk oplossen van het probleem, een passende niveauverhogende activiteit kan zijn. Dit is het geval wanneer er op het werkblad gevraagd wordt om de stappen voor het vinden van de oplossing van een probleem individueel volledig uit te werken en daarop te reflecteren. Om meer uitgebreide betekenisvolle reflecties van leerlingen te stimuleren blijkt dat het afhankelijk is van de vordering van een groep leerlingen binnen het proces van het probleemoplossen of een docentgestuurd klassengesprek of een leerlinggestuurd groepsgesprek relevanter is. Een bijzonder resultaat hierbij is dat wanneer leerlingen het centrale probleem uit een les al snel hadden opgelost, zij in sommige gevallen uit zichzelf met elkaar gaan discussiëren over hoe de gevonden oplossing toegepast of geoptimaliseerd kan worden en dit vervolgens gaan uitvoeren.

Vervolgonderzoek zou zich kunnen richten op de invloed van deze lessen op de leerprestaties van de deelnemende leerlingen binnen het curriculum. Hierbij krijgen namelijk wiskundige kennis en onderwerpspecifieke heuristieken, de overige twee essentiële bekwaamheden die Alan Schoenfeld beschrijft voor het wiskundig probleemoplossen, meer de nadruk.

Erasmus MC is conducting a study on the increased risk of cardiovascular disease in women who have migraine. For this purpose, 50 healthy women and 50 women with migraine in age category of 40-60 years were compared using measurements of their dermal blood flow.
These measurements are done using two different imaging devices: Laser Speckle Imaging and Laser Doppler Imaging. The result is a recording of the fluctuations in the dermal blood flow over a time period influenced by circumstances and various mechanisms and consists of signal like curves. Erasmus MC have processed these measurements and obtained results. However, they are interested in whether applying Fourier Transform could offer additional insights or new (perhaps faster) methods to process the measurements.
The signal like curves can possibly be decomposed into sine and cosine waves using Fourier Transform. The result of the Fourier Transform consists of a visualisation of the intensity of the various components of such a waveform/curve. Based on that, it is possible to estimate the contribution of the mechanisms to blood perfusion. Existing literature and previous studies provide frequency intervals related to several mechanisms that influence the dermal blood flow.
This thesis will research to what extent the mechanisms contribute to the dermal blood flow across the different phases and how their behaviour differs when the circumstances for the measurement are adjusted. Perhaps there are behaviours that are instantly recognizable across all patients. An interesting question to answer would be whether it is possible to directly visualise the risk groups.

All organisms are built out of cellular tissue. Being able to recognise abnormalities in these tissues could be useful in recognizing cancerous cells. In this thesis we construct a mathematical model for cellular tissue based on its spatial structure. We consider cells as elements of the network. Touching cells are considered connected. Cells grow at different growth speeds. We determine the point in time when this network is fully connected, meaning there is a path between every pair of cells through touching cells. This point indicates the start of the last phase of the cellular growth, where friction restricts cell movement. We first use a Poisson point process to generate the locations of the cells. To make the model more similar to cellular tissue, we introduce determinantal point processes which have short-ranged repulsion, meaning points repel each other and thus spread. We compare the repulsion of Poisson and Determinantal point processes with real cellular tissue. We conclude that determinantal point processes have significantly higher repulsion than Poisson point processes. We also conclude that the repulsion in cellular tissue is higher than both point processes. Using simulations, we show that the model with determinantal point processes reaches connectivity significantly earlier than the Poisson model. We conclude that the analytically derived connectivity time point from the Poisson model can be used as an upper bound for the determinantal model.

This research is conducted in collaboration with the Sophia Children's Hospital (SCH). The hospital wants to provide their patients with more detailed information about when a patient is approximately scheduled to have a surgery. The first step is to create a model which optimises the operation room (OR) schedule and indicates when different kinds of surgeries are planned. This information, combined with the waiting list, provides insight in when a surgery of a specific patient is scheduled.

In a hospital, different departments work together to treat the patient as good and efficient as possible. If a patient needs a surgery, not only an OR is needed, but also a bed at a ward which matches the patient's needs. The goal of this thesis is to use the different resources of the hospital as efficiently as possible. This is done by not only optimising the utilisation of the OR, but at the same time levelling the bed occupancy of the different wards. The levelling of the bed occupancy is done by minimising the maximum number of used beds at each ward. Because, if we minimise the maximum, we force that the patients are spread out more evenly over the day.

For each specialty, the patients are divided into patient groups based on historical data using a constrained $k$-means clustering algorithm. For each patient group, information is gathered about the length of stay (LoS) and the surgery duration of patients in this patient group. Next to that, the number of patients in a patient group indicates how often a patient group needs to be scheduled at least.

The probability distribution of the surgery duration is taken into account when deciding at which day, at what time, and in which OR a surgery is planned. A patient group can only be scheduled during OR shifts assigned to the corresponding specialty. At the same time, the levelling of the bed occupancy is taken into account.

After some constraints are linearised, this model can be formulated as a mixed integer linear program (MILP). However, the model has a large number of variables. Therefore, column generation is used to split the model into smaller subproblems per specialty. Some of the pricing subproblems take a lot of time to optimise. For that reason, we set some time limits both on the runtime of the pricing subproblems and the runtime of the entire algorithm. Column generation does not guarantee an optimal solution of our MILP. However, the objective value of our MILP improves over time, when new columns are added to the set of available columns. This indicates that column generation can be used to optimise our model.

In this thesis, several versions of the model are presented. For example, the schedule is different if the bed occupancy is calculated every hour or of every fifteen minutes. Next to that, the model can either be more focussed on maximising the OR utilisation or on levelling the bed occupancy.

The problem considered in this thesis is the box set optimization problem. In
this problem the goal is to find the optimal dimensions of a number of shipping
boxes which have to provide an optimal fit on a set of items. The application
considered is e-commerce, and hence the number of items is relatively large
compared to other applications found in literature.

The literature describes multiple problem formulations where a small num-
ber of optimal shipping boxes are selected from a large set containing candidate
boxes. However, since the instances considered in e-commerce application are
often too large to be solved in this way, we introduce a simulated annealing
heuristic solution method. In this way we are able to generate box sets con-
taining 10 to 20 boxes with centimeter accuracy for data sets containing up to
XX single item orders.

We pay special attention to an algorithm to very efficiently calculate the
objective value associated to the fit of a selection of candidate boxes on a set of
items. This is done by assigning items to shipping boxes using a pre-calculation
table. Because the heuristic can process a larger set of candidate boxes than
the exact solution method, ultimately results of the heuristic are better than
those found by the exact solution method.

Finally we generate multiple box sets based on the input data provided by
the industry partner bol.com, and compare the results. We conclude that on
almost all tested instances, the total amount of carton used can be reduced by
more than XX in combination with a total send volume reduction of over XX
in comparison with the box set that is currently used. Furthermore, we answer
additional questions. We find that minimizing the total amount of carton used,
instead of minimizing the total volume send, can lead to a result that is most
desirable in practice. We discover that pack-line specific, warehouse specific or
seasonal specific box sets provide none to only minor additional improvements.

A big proportion of hospital-associated infections caused by Staphylococcus aureus can be attributed to Methicillin-resistant S. aureus (MRSA). Many countries take interventions to try and minimise the spread of MRSA. Interventions, such as a search-and-destroy policy and restrictive antibiotics use, have proven to be effective. Different strains of MRSA are grouped into clonal complexes (CCs) by their similarity to a central allelic profile. Separate MRSA CCs have evolved independently over time. In most countries a limited number of CCs is responsible for the prevalence in the population. In each country different CCs are present and the exact reasons why these CCs are successful in the specific countries is unknown. The aim of this project was therefore to study two different models, one implemented in R and one implemented in Java, that both simulate the spread of multiple MRSA CCs in a population. The two individual based models (IBMs) considered in this research produce similar results when setting them side-by-side using simple model set-ups. The R model turned out to be computationally expensive and very restrictive, where on the other hand, the Java model was much faster and more extensive. Consequently, the Java model was used to simulate the spread of MRSA CCs in a more advanced setting. Although the model set-up and population parameters were not yet realistic, it demonstrated some interesting findings. The general observation was that CCs with higher antibiotic resistance contribute most of the MRSA infections. The model showed that intervention by means of a search-and-destroy policy can lower the overall prevalence in a population significantly and create more variation between the CCs present. Since the model set-ups adopted during this research most likely do not completely agree with the biological processes, populations and interventions in the real world, future research should determine whether the obtained exploratory results also hold true in more representative populations.

This research is mainly about optimising compartment models for COVID-19 and then using them for different applications that can be used to consult authorities. A compartment model describes the dynamics of a disease by implementing differential equations for the different states that belong to that disease. This is done using parameters which can be estimated from actual data or can be extracted from researches done by external authorities. We begin by extending the standard compartment model to a model which is applicable to COVID-19, this extension is based on characteristics of the virus. After this extension we are going to counteract on the assumption that susceptible and infected individuals are heterogeneously in contact with each other, since social distancing and quarantine prevent these two compartments from interacting. Also, we use specific time intervals and optimise the mortality rate. After these improvements we create an external model which estimates a contact tracing queue such that authorities can forecast how many contact tracers are needed to keep the pandemic under control. Then using the implementation of this queue, another queue is made which is dependent on its own length to forecast what happens if the government is not in contact with individuals who possibly are infected with COVID-19. Then we extend the model in such a way that vaccinated individuals can be assigned to a specific compartment, and at last we create a queue as described above with this extended model.

This thesis is devoted to option pricing on backward-looking rates. For the last decades, interest rate products were often linked to IBOR rates. IBORs are short-term borrowing rates charged between global banks in the unsecured interbank market. The purpose of this thesis is to compare the Hull-White model to the Black-Karasinski model for the pricing of caps/floors on compounded rates. Both models are so-called short-rate models, which are widely used for interest rate modelling. Due to the IBOR reform, new products are expected to appear in the market. One type of these products is caps/floors linked to the new Risk-Free Rate (RFR). The new RFRs will be in-arrears backward-looking rates and, as a consequence, have an impact on the choice of pricing models.
This thesis considers caps/floors on the new compounded RFR rates. For both models various pricing techniques for caps/floors on compounded rates are investigated. For the Hull-White model, the pricing kernel approach and a Monte Carlo simulation are explored. The pricing kernel approach yields an analytic formula for caps/floors on compounded rates. This formula is also used for the comparison. For the Black-Karasinski model, the pricing kernel approach, the trinomial tree method and the Monte Carlo simulation are considered. The pricing kernel approach yields a semi-analytic formula for caps/floors on compounded rates. However, in practice the computation time of this semi-analytic formula turned out to be substantial. Further, despite the fast computation time of the trinomial tree for LIBOR caps/floors, the trinomial tree method is rather slow for the caps/floors on compounded rates. As a result, the Monte Carlo simulation is the most suitable pricing technique, along the three
explored methods, for caps/floors on compounded rates under the Black-Karasinski model. Therefore, the Monte Carlo simulation is used for pricing caps on compounded rates in the model comparison.
Since there exists no liquid market yet for caps/floors linked to the new RFR, the models are calibrated to a proxy market. First, the Black-Karasinski model is calibrated to the proxy market using a heuristic approach. This heuristic approach is chosen as a compromise between computation time and accuracy. Then, the Hull-White model is calibrated to the Black-Karasinski model using the stripping method. Having both models calibrated, the price of caplets/floorlets on compounded rates are calculated. Thereafter, these prices are inverted to Bachelier implied volatilities for a uniform comparison. From the comparison of the two models, a difference in Bachelier implied volatility is observed in a range of 4 to 4 bps. This is of one order less than the volatility itself.

Thyroid hormones are fundamental in the development and function of the human body. They are produced by the thyroid and their concentration is regulated by a negative feedback loop involving also the hypothalamus and the pituitary. Many studies have shown that each individual has a unique HPT axis set-point, which means that everyone has his or her own personal level of thyroid hormones. However, when the thyroid is affected by a disease, such as hypothyroidism, a change in the thyroid function modifies the levels of thyroid hormones, therefore they do not match the patient’s set-point anymore. Hence, the aim of this thesis is to investigate the relationship between the thyroid hormones TSH and FT4 in order to predict each individual’s set-point. Moreover, the variation in time of the two hormones in patients under medication is studied as well, which allows to determine around which values the concentrations of TSH and FT4 will stabilize and, if these are not matching the set-point, to adjust the patient’s medication dosage. This research was conducted using patients’ datasets from different hospitals. In particular, datasets of hypothyroid and thyroidectomized patients were included in this study. The analysis of hypothyroid patients’ measurements shows that TSH can be expressed as a negative exponential function of FT4. Furthermore, the individual set-point can be predicted as the point of maximum curvature of that function and this has been validated using data from thyroidectomized patients. Finally, a model of coupled differential equations, taking into account the effects of medication, has been developed to describe how the concentrations of the hormones are changing in time after a patient starts the treatment.

Cholera is an infectious disease caused by the bacterium Vibrio cholerae. Infections can occur directly via contact with infected people, but also indirectly via infected water. Because of these direct, but especially indirect infections, it is useful to investigate the spread of the disease using mathematical epidemic models. In this report a compartmental epidemic model for cholera is discussed: the SIRB model. This model consists of four compartments: susceptibles, infected, recovered and the bacterial concentration in water. The interaction between the groups is given by a system of differential equations. The equilibrium solutions and their stabilityare mathematically analysed. The main focus of this report, however, is on two different spatial models. The patches model consists of multiple patches connected by migration of people. Different measures against cholera are studied using this model. The second spatial model that has been numerically implemented is a diffusion SIRB model. The influence of the diffusion coefficients on the solution patterns will be researched for different initial conditions. The results presented in this report are building blocks for further research into compartmental models that simulate the spatial dynamics of cholera.

Dynamic Time Slot Management (DTSM) is a system often used in online retail to manage the delivery of goods to customers. With DTSM customers arrive over time and place orders. They get presented with a set of time slots and the customer picks the time slot in which he wants the goods to be delivered to his home. A DTSM system creates a time slot offer for the customers based on the current delivery schedule, which is a solution to a Vehicle Routing Problem (VRP). Each accepted customer is added to this delivery schedule. This delivery schedule gets periodically optimized by an optimization algorithm. During this optimization, new customers may arrive and receive time slots based on the non-optimized schedule, which causes a discontinuity in the system. In this thesis, five different procedures have been created that deal with this problem. The impact of the different procedures is analyzed by simulating the DTSM system and comparing the final delivery schedules. Solving the problem by leaving out the optimization or insertion step of the DTSM system is outperformed by all procedures. Procedures that barely make use of the optimized schedule accept the least number of customers. The procedure that delays customers, which makes it similar to the theoretical setting, accepts the most customers, but this comes at the cost of poor customer service. The most promising results are found by the procedure that inserts new customers into the optimized schedule. Enhancing this procedure with a merge algorithm improves the performance slightly.

In dit verslag beschrijven we met behulp van een wiskundig model hoe een malaria infectie zich ontwikkelt in een persoon. We gebruiken hiervoor een stelsel van gewone differentiaalvergelijkingen, wat beschreven wordt in een artikel van Mary Bushman. Hierbij volgen we veranderingen in verschillende aantallen die een rol spelen bij een infectie, zoals het aantal geïnfecteerde en ongeínfecteerde rode bloedcellen, het aantal parasieten en de immuniteit, die wordt opgedeeld in een aangeboren deel en een adaptief deel. Hierbij kijken we naar de invloed van medicijngebruik en de infectiehistorie op een infectie. We gebruiken een numerieke rekenmethode om het stelsel op te lossen. We gebruiken de functie ’ode45’ in het programma MATLAB om tot een oplossing te komen. Bij een infectie door zowel vatbare parasieten sterven de vatbare parasieten eerder uit. Dit heeft te maken met de ontwikkeling van de adaptieve immuniteit tegen een bepaald soort parasiet. Het toedienen van medicijn vroeg in de infectie heeft meer effect dan later. Als laatste hebben we gekeken naar de invloed van eerdere infecties op de ontwikkeling van een nieuwe infectie. Doordat het immuunsysteem een soort parasiet al eerder is tegengekomen, geeft dit een nadeel voor de parasiet. Dit zorgt ervoor dat de parasieten sneller uit zullen sterven.

Currently, speaker recognition research is mainly based on phonetics and speech signal processing. This research addresses speaker recognition from a new perspective, analysing the transcription of a fragment of speech with text analysis methods. Since text analysis is based on the transcription text only, it can be assumed independent from current automatic speaker recognition software. Hence, it would contribute significantly to the overall evidential value. The analysis is based on the frequencies of non-content, highly frequent words. We study whether information about the identity of the speaker is contained in the transcription of spoken text. The value of evidence is quantified using a score-based likelihood ratio. The score-based approach is chosen because in most forensic cases, there is not enough data from the suspect or of the disputed speech fragment available to model a robust feature-based likelihood ratio. Different methods to model the system from feature vector over score to likelihood ratio have been compared. As a baseline, a distance based method is used, where the score is the distance between the feature vectors. To improve upon this baseline, machine learning algorithms are implemented. The results from SVM and XGBoost are explored. As a third method a feature-based likelihood ratio is calculated and used as a score instead of as a direct likelihood ratio. With this method, both similarity and typicality are taken into account. The model is trained and tested on the FRIDA data set from the Netherlands Forensic Institute, consisting of Dutch conversations from a homogeneous group of 250 individuals. The performance of the likelihood ratio system is evaluated through computing the cost log-likelihood-ratio (Cllr), which is a measure for the accuracy and quality of the likelihood ratios, and the accuracy (A) of the likelihood ratios solely. The performance is also evaluated by inspecting the Tippett, empirical cross-entropy and pool-adjacent-violators plots. Different values for parameters used in the calculation of the likelihood ratios are investigated: the length of the sample, the number of frequent words (number of features) and the number of samples needed to train the model. The distance method showed a strong baseline, with good performance for large sample lengths. The SVM method outperformed the distance method for all parameter settings, with a peak performance of A=0.94 and Cllr=0.24. The XGBoost method showed promising results for smaller samples lengths, but a too large amount of data is needed to obtain good performance for larger sample lengths. The LR score method showed moderate results, but no improvements due to the necessity to estimate high-dimensional distributions. This thesis shows that information about the identity of the speaker is contained in transcriptions of speech. The complete process from data to likelihood ratio is constructed, where the likelihood ratio quantifies the evidential value of a transcribed speech fragment.

Combined chemo-radiation treatments are often used to enhance tumor cell kill, with respect to their separate treatments. Currently, chemo-radiation therapy treatments are prescribed as a separately optimized radiotherapy treatment with the addition of a generic chemotherapy regimen. The dosing of the combined therapies is mostly based on clinical experience. Hence there is a clear need for the optimization of dosing for combined chemo-radiation therapy treatments. The goal of this thesis is to propose novel approaches to optimize cancer treatment efficiency when combining chemotherapy and radiotherapy by using control-theoretical methods as well as radiation therapy models. A positive switched system is a dynamic system consisting of a set of Metzler state-space matrices and a specific switching law, which determines when and how to switch between the subsystems. Positive switched systems seem particularly appropriate to model cancer evolution under different treatments, and therefore to determine optimal treatment scheduling. They are used to model the effect of combined chemo-radiation treatments, thus enabling the systematic design of optimal therapy planning. The proposed model is especially appropriate for heterogeneous tumors, since it describes the impact of therapies on different tumor cell lines, and includes the possibility of mutations. Metastases are also included since the model considers separate tumor compartments, and the possibility of migrations between compartments are modeled as well. Undesired consequences due to toxicities are incorporated, and an upper bound can be set to limit the possible damage. A clinical dataset was obtained from the Erasmus MC, which was used to validate chemotherapy-alone model outcomes. Combined chemo-radiation treatments were also developed for a set of initial tumors. The model shows promising applications for cancer treatment design. The use of positive switched systems for combined chemo-radiation treatments is a new approach to obtain optimal drug usage and radiation fractionation schemes for each body compartment. The achieved findings could have a significant impact on effective treatment planning by choosing optimal drug durations or by using radiotherapy in combination with chemotherapy, for instance, to target certain chemotherapy-resistant areas, to achieve tumor control with minimal side effects.

Optimizing supply chain management in hospitals can lead to a significant reduction in costs. One of the supply chain processes that can be optimized is the sterilization of reusable instruments used during surgical procedures. This process takes place at the Central Sterile Supply Department (CSSD) and consists of two main steps, `washing and disinfection' and `sterilization'. These steps are executed by batch processing machines, with a setup time prior to each step. The setup at each stage is executed manually. The machines of a CSSD are renewed every 10 years, which provides a opportunity to assess all capacity planning decisions. In this thesis, the aim is to determine the required resources of the CSSD, while minimizing the total costs and guarantying the availability of instruments for scheduled surgeries. Costs include: acquiring machines, machine batch costs, machine maintenance costs, and staff costs depending on the number of opening hours. This thesis contributes to current research by proposing a framework for the capacity planning decisions at a CSSD, by extending existing models by taking specific characteristics of the CSSD into account, and considering a new objective function, namely minimizing the total costs. Capacity planning decisions are considered on three hierarchical levels. First, the strategic level involves long-term decisions, where the number and type of required machines have to be determined. Second, on a tactical level, the amount of opening time has to be determined. Third, on an operational level, the instrument sets have to be scheduled within batches and machines. The sterilization process is formulated as a mixed integer linear problem (MILP). The problem is described as a multi-stage hybrid flow shop with additional constraints to represent the specific characteristics of the CSSD. This model takes into account capacity planning decisions on all three levels. The MILP formulation is proven to be NP-hard. The demand for sterilized instruments is determined from historical data from the Leiden University Medical Center. To evaluate the model, three instances with a timespan of a week are created. Preliminary results showed that the multi-stage flow shop is difficult to solve for real-life instances. Hence, a decomposition, based on the three hierarchical levels, is proposed. The decomposition leads to three levels of optimization models. The strategic model takes into account the scheduling of instruments per day, while the tactical model schedules per day parts, and the operational models determine a specific point in time. The proposed models are individually tested on their performance. The results from the strategic model are used as input for the tactical model. Results show that the strategic model underestimated the required amount of opening time, as it does not take into account the spread of release times. Furthermore, the computational results show that the tactical and operational models are difficult to solve for real-life instances. Hence, a heuristic approach is proposed by forming a chain of the models. The best results were obtained by setting a minimum amount of opening time for the strategic model and using the resulting machines as input for the tactical model. The results show that the instrument sets are equally spread over the week. To conclude, the results of this thesis contribute to quantifying the required resources for a sterilization process within a hospital. However, to obtain more practical results, future research is required. Suggestions for future research include: taking uncertainties within the process into account, the application of metaheuristics, and the required number of employees.

De CT-scanner heeft een belangrijke rol in de medische wereld. Helaas gebruikt de CT-scanner voor het opstellen van een scan schadelijke straling. In dit onderzoek is nagegaan of er een betere beeldreconstructietechniek te vinden is voor het opstellen van een CT-afbeelding met gebruik van minder straling.
De standaard voor het opstellen van een CT-afbeelding is het gebruik van de 2-norm. Vergelijken we de CT-afbeeldingen geconstrueerd met de 2-norm met CT-afbeeldingen geconstrueerd met de 1-norm en 0.5-norm dan zien we dat de 1-norm visueel het beste resultaat oplevert.
Naast het vergelijken van de verschillende normen hebben we ook verschillende epsilon-regularisatiemethodes bekeken. De numerieke experimenten met de verschillende epsilon-regularisatiemethodes hebben we uitgevoerd op een systeem met 0%, 3% en 10% ruis, om te onderzoeken welke methode het beste toepasbaar is in de praktijk. Van de door ons geteste epsilon-regularisatiemethodes bleek de epsilon-regularisatie met epsilon := epsilon/2 met startwaarde epsilon = 10^(-8) en gebruik van de 1-norm de meest geschikte methode om een onderbepaald CT-probleem op te lossen.