There is increasingly more expensive maintenance that needs to be performed on the Dutch railway network. Good maintenance schedules reduce costs, minimise hindrance to passenger and freight travel, and follow restrictions imposed by available resources, legislation and other agreements. The railway maintainer has modelled this maintenance scheduling problem, among others, on an annual level. This problem definition is very precise with different conflicting non-linear constraints and cost parts. The demands of the solving method depend on the phase of the scheduling process. In early phases, low runtime is most important, whereas best solution quality outweighs this runtime when the maintenance work is more finalised. In 2019, a simple greedy algorithm found good solutions fast, and a hybrid greedy-evolutionary algorithm was developed that resulted in the best schedules. However, since then, the problem definition has been made more realistic and thus complex. Therefore, this hybrid greedy-evolutionary algorithm is no longer feasible, and creating a maintenance schedule takes significantly longer than before. Research is necessary to better understand the impact of the more realistic model, and to once more have a good trade-off between solving time and solution quality available for the schedulers. In this thesis, we aim to achieve this by improving different aspects of the problem and solving methods. Most experiments were done with the maintenance schedule of 2024, and results were verified on the years of 2023 and 2025. First, general problem analysis and implementation improvements reduced the runtime from around twenty-four hours to three hours with the greedy algorithm. Then, approximations were applied in the passenger hinder to further reduce the runtime by around half with a negligible negative impact on the solution quality. Due to these speed-ups, it was possible to use more elaborate solving methods. New experimental results showed that the greedy algorithm still finds solutions fast. The hybrid greedy-evolutionary algorithm found better quality schedules, but required more runtime. Furthermore, a novel solving method with look-aheads was proposed, which showed some potential for cost reductions, but was dominated by the hybrid algorithm. Every algorithm uses the greedy heuristic as a subroutine. Results showed the importance of finding the right order for greedily scheduling the requests. A proposed new order function improved the quality of the resulting maintenance schedules even further. To conclude, the increase in complexity of the problem definition in recent years has made solving more difficult. To still find good solutions in a similar time, a better performing greedy heuristic was necessary. By applying different runtime optimisations to the objective evaluation and improving the solution quality of the greedy heuristic, a good trade-off between runtime and solution quality, using different solving methods, was realised for creating annual maintenance schedules for the Dutch railway network.

During a viral infection, we expel remnants of the virus. This makes it possible to conduct wastewater analysis which aid in the efforts to track the evolution of the current Covid-19 pandemic. It has been shown that by repurposing the kallisto algorithm, the abundance of SARS-CoV-2 variants in wastewater samples can be estimated. Since this is a novel method for this scope, its precision could probably be improved by adjusting certain aspects. In this work, I look at one of those aspects: sequencing particular genomic regions of the virus rather than the entire genome. I have indeed found that the regions that code for the spike (S) and nucleocapsid (N) regions and a section around the region coding for the non-structural protein 3 (nsp3) give particularly accurate results when sequenced on their own. In addition, in at least one case, combining two well-performing regions further improves accuracy at lower simulated abundances of variants. This suggests that sequencing depth is preferred over sequencing breath as long as the region being sequenced contains enough information to distinguish between variants. These findings are important as they can aid in the improvement of this method of variant quantification. Moreover, they can also help in improving other algorithms applied to the SARS-CoV-2 genome by highlighting the genomic sections containing the most differentiating information between variants.

AmpliDiff is an algorithm for studying environmental samples. As the study of those DNA samples is complicated, the runtime of AmpliDiff holds back its usability. One time-consuming part is exactly solving a variant of the Set Cover problem. This paper researches whether this exact solution can be replaced by heuristics to reduce the runtime. The Iterated Local Search framework is applied as heuristic and translated onto our problem. After that, it is tested and compared to the original version of the algorithm based on the found solution size. From the results we can deduce that the heuristic version finds the same solutions as the original. It does so in a faster relative time frame as well. However, due to time constraints, the dataset that was used for testing is rather small and therefore not representative of more complex problems, which are usually solved by AmpliDiff. Thus, for future work we recommend to test the heuristics version on a larger, more representative dataset to check its viability. This paper has laid the groundwork by implementing the heuristic and showing that it effectively works for small datasets, giving reason to also try it out on more complicated problems.

Positive end-expiratory pressure (PEEP) is one of the components of mechanical ventilation treatment for patients with acute respiratory distress syndrome (ARDS). Correct PEEP level can reduce additional lung injuries sustained during the hospitalisation, significantly increasing patients' chances for survival. In this paper, we focus on estimating the difference in patient mortality when assigned high or low PEEP level. We look at three machine learning models specifically designed for such tasks: S-learner, T-learner and causal forest. Through a series of experiments, we determine their best use cases based on simulated data and measure their performance on a real-life dataset - MIMIC-IV. In our analysis, we find that after tuning the hyperparameters, the models can, to some degree, make valuable predictions and reveal heterogeneity in the treatment effect. However, when evaluated on a separate dataset, the models' performance drops significantly.

Mechanical ventilation with positive end-expiratory pressure (PEEP) is a critical intervention for patients in intensive care units (ICUs) with acute respiratory failure. Identifying the optimal PEEP level is challenging due to conflicting evidence from studies comparing low and high PEEP regimes. This research explores machine learning methods for estimating individualized treatment effects (ITE) in ICU patients on different PEEP levels using the observational MIMIC-IV dataset. Various conditional average treatment effect (CATE) estimators, including S-, T-, and DR-learners, are applied to control for confounders and identify PEEP effects on patient subgroups. This research aims to compare the performance of the aforementioned CATE estimators, with a focus on the doubly-robust (DR) learner, and determine which one is best suited for causal inference in this context. The DR-learner offers increased resilience to model errors since it integrates two models. Simulations using mean squared error (MSE) show the DR-learner performs well with confounded data and differing linear response functions between control and treatment groups. However, when looking at the performance on the MIMIC-IV dataset, the predictions are unstable, failing to reliably identify the optimal PEEP for increasing patient survival. This trend is also observed in a randomized controlled trial (RCT) dataset, with the area under the Qini curve (AUQC) close to zero, indicating difficulties in identifying the effects of PEEP settings. Despite promising simulation results, real-world application shows limitations in these machine learning methods for optimal PEEP identification.

Mechanical ventilation with positive end-expiratory pressure (PEEP) is a critical intervention for patients in intensive care units (ICUs) with acute respiratory failure. Identifying the optimal PEEP level is challenging due to conflicting evidence from studies comparing low and high PEEP regimes. This research explores machine learning methods for estimating individualized treatment effects (ITE) in ICU patients on different PEEP levels using the observational MIMIC-IV dataset. Various conditional average treatment effect (CATE) estimators, including S-, T-, and DR-learners, are applied to control for confounders and identify PEEP effects on patient subgroups. This research aims to compare the performance of the aforementioned CATE estimators, with a focus on the doubly-robust (DR) learner, and determine which one is best suited for causal inference in this context. The DR-learner offers increased resilience to model errors since it integrates two models. Simulations using mean squared error (MSE) show the DR-learner performs well with confounded data and differing linear response functions between control and treatment groups. However, when looking at the performance on the MIMIC-IV dataset, the predictions are unstable, failing to reliably identify the optimal PEEP for increasing patient survival. This trend is also observed in a randomized controlled trial (RCT) dataset, with the area under the Qini curve (AUQC) close to zero, indicating difficulties in identifying the effects of PEEP settings. Despite promising simulation results, real-world application shows limitations in these machine learning methods for optimal PEEP identification.

In the intensive care unit (ICU), optimizing mechanical ventilation settings, particularly the positive end-expiratory pressure (PEEP), is crucial for patient survival. This paper investigates the application of neural network-based machine learning methods to personalize PEEP settings in the ICU, aiming to improve patient survival outcomes. The research focuses on two specific algorithms, TARNet and CFR, evaluating their ability to estimate individualized treatment effects of lower versus higher PEEP regimes. The study is structured into three phases: controlled simulations, application to the MIMIC-IV dataset, and validation using a randomized control trial dataset. TARNet and CFR showed potential for estimating the individualized treatment effects but required large datasets for optimal performance. In the case where limited data is available, these models are upstaged by simpler learners, such as the S- and T-learners. The study concludes that while neural network-based methods hold promise for personalizing ICU treatment, their efficacy is heavily influenced by data availability and quality.

This research investigates the use of Causal Multi-task Gaussian Process (CMGP) for estimating the individualized treatment effect (ITE) of low versus high Positive End-Expiratory Pressure (PEEP) regimes on ICU patients requiring mechanical ventilation. The study addresses the complexities of determining ITE due to the inability to observe counterfactual outcomes and the confounding bias in observational studies. By employing Conditional Average Treatment Effect (CATE) estimators, such as S-Learner, T-Learner, and CMGP, the research evaluates the impact of different PEEP settings on patient survival across varied patient characteristics. The precision of these estimators is assessed using simulated data, real-world observational data from the MIMIC-IV dataset, and an external RCT dataset. The findings of this study are inconclusive, highlighting the need for further research to refine these methods and explore larger, more balanced datasets.

Mechanical ventilation is a vital supportive measure for patients with acute respiratory distress syndrome (ARDS) in the intensive care unit. An important setting in the ventilator is the positive end-expiratory pressure (PEEP), which can reduce lung stress but may also cause harmful side effects. This research investigates the personalization of PEEP settings based on patient characteristics using three meta-learning algorithms (S-, T-, and X-learner) to estimate the conditional average treatment effect. Additionally, the hypothesis that the X-learner performs particularly well under a significant imbalance in patient numbers between treatment groups is tested. Results show that the X-learner slightly outperforms the S- and T-learners in terms of mean squared error under various unbalanced conditions in simulated data. However, the overall ability of these meta-learners to identify patients benefiting from high PEEP remains inconclusive. When using gradient boosted trees or random forest as base models, cumulative gain curves on MIMIC-IV data indicate potential overfitting. While the X-learner performs somewhat better on this data, the low area under the curve scores suggests a minimal distinction between high and low PEEP groups. External validation with data from a randomized control confirms that the models do not effectively distinguish between treatment groups. These findings suggest that further investigation with more complex models and real-world data is needed to validate the potential of meta-learning algorithms in personalizing PEEP settings for ARDS patients.

AmpliDiff provides a method which takes a list of genomes and their lineages, and finds a set of amplicons and their primers in such a way that these amplicons can be used to differentiate between the lineages of a specific virus. While it has been shown that AmpliDiff find results comparable to whole genome sequencing for SARS-CoV-2 when looking at abundance estimations, it is not know how well it performs for other viruses, or what factors of a virus impacts the performance of the amplicons found by AmpliDiff.\ In this paper we will be showing the effectiveness of AmpliDiff on Human monkeypox, HIV-1 and Influenza-A. By running AmpliDiff for the three viruses mentioned above, we obtain sets of amplicons, which are used to do a lineage abundance estimation. By then comparing the estimation to the know abundance we calculate the Mean Average Error (MAE). This MAE will then be used to compare against the MAE obtained from doing a abundance estimation based on whole genome sequencing. By comparing the amplicons against whole genome sequencing (wgs), we show that using viruses with longer genomes positively impacts the performance of the amplicons. We also show that the amount of misalignment characters added by the Multiple Sequence Alignemnt (MSA), impacts the required settings for AmpliDiff to find amplicons, and can negatively impact the MAE. Finally, we show that AmpliDiff can be run, with some minor changes to the code base, on segmented genomes, with performance similar to that of single segment genomes.

The human microbiome, the ensemble of microorganisms found in and on the human body, plays a key role in human health and disease. However, the current state of microbiome analysis represents a significant challenge for machine learning algorithms. Datasets of microbiome sequences are often characterized by a regime of large dimensionality and relatively few labels, making it difficult for a model to discriminate features from random noise and avoid overfitting. It is, therefore, paramount to reduce the dimensionality of the input data while preserving their structure and information for a model to properly learn from them. K-mer frequency vectors and learnable representations through encoders are some of the embedding methods that have been proposed in literature to reduce the dimensions of the input space for machine learning algorithms operating on biological sequences. This work aims to compare how various embedding techniques influence the performance of a downstream disease detection task from microbiome sequencing data. In particular, the research shows that k-mer frequency vectors lead to better classification metrics (AUC = 0.88) compared to NeuroSEED embeddings (AUC = 0.76) on euclidean space. The work also presents how the classification problem formulation is critical to improving the overall disease detection performance.

Since the start of the SARS-CoV-2 pandemic, the monitoring of SARS-CoV-2 by way of viral RNA sequencing of wastewater has proven to be an efficient and effective way of estimating COVID-19 cases in population groups. A recently developed pipeline also enables us to estimate SARS-CoV-2 variant abundance using viral samples from wastewater. This is done by repurposing an RNA-seq quantification algorithm to quantify reads, belonging to variants, from DNA-sequencing data. However, the impact of sequencing errors and contaminating viruses on this process is unknown. Here I show that, in simulated data, the credibility of the prediction results is dependent on the error rate of the sequencing machines used. I also show that contaminating the simulated dataset with certain human coronaviruses has a significant effect on prediction accuracy. However, most viruses currently found in wastewater have no effect. Furthermore, adding a reference genome for these human corona-viruses to the reference set removes any impact. The results demonstrate that it is important to assess the credibility of the pipeline on a case by case basis and to tailor the testing setup and reference set to this assessment.

With widespread use of advanced technology for the recording, storing and sharing of social interactions, protecting privacy of people has been a growing concern. This paper zooms in on the collection of spoken audio with regard for the privacy of recorded individuals. Recently efforts have been made to collect audio at a low sampling rate to obfuscate spoken words in the recorded audio, such that conversations are kept private. This research investigates whether it is possible to upsample this low-resolution audio, using an existing super-resolution model, in order to reveal parts of the previously obfuscated conversations. The performance of the model is measured in terms of the word error rate of automatically generated transcriptions of the upsampled audio. It turns out that it is possible to significantly increase the intelligibility of low resolution privacy-sensitive audio by upsampling. Though the use of the super-resolution model seems to be limited when it comes to revealing significant parts of conversations.

The interactions between human and machines are now common in our daily life. The audio data of human communication is a rich source of information, but it is con- sidered privacy-invasive for machines to listen to it. By reducing sampling frequency, it is possible to preserve privacy by making conversation unclear while still being possible to detect if someone is speaking or not. The topic of this paper is to investigate how low sampled frequency audio data hinders the detection of speech. To detect speaking, voice activity detection has been applied, which is a technology in the signal process- ing field that identifies which short segments of audio contain speakings. Two types of state-of-art voice activity detector(VAD) were used for this experiment including a supervised (pyannote) and two unsupervised (rVAD pitch and flatness mode) methods. As a result, the unsupervised methods outperformed the supervised model, where rVAD pitch mode has resulted in the best performance out of all three. More specifically, the unsupervised VAD’s performance became lower as the sample rates decreased while the supervised VAD did not work well at higher sample frequency. rVAD pitch mode at sample rates of 8000Hz or higher was possible to perform at the almost same level as a state-of-art supervised VAD that is trained in a similar data set. Furthermore, it was able to perform as well as a modern unsupervised VAD at 2000Hz or higher sample frequencies. At the sample rate of 1250Hz or lower, any VAD was not able to perform at the same level as a state-of-art VAD. Regarding the privacy aspect, it is observed that human ears detect speaking better than computers, where humans can understand parts or all of the contents of speaking at 2000Hz or higher, which infers that current technology is not enough to detect speech from downsampled privacy-preserving audio. However, there is still a need for further research to verify the effects of the training set and its sample frequencies for the supervised method and also proper scientific so- cial experiments to test the ability of humans of speech detection for reduced sampled audio.