An accurate segmentation model for hip compo- nents could improve the diagnosis of Osteoarthritis, a prevalent age-related condition affecting joints. A significant challenge in developing effective and robust segmentation models are the domain differ- ences across various datasets. In this study, we in- vestigate the impact of different data augmentation and preprocessing techniques on the generalizabil- ity of femur segmentation models across datasets. Using two labeled datasets, we evaluate the perfor- mance of a U-Net segmentation model, focusing on the effectiveness of augmentations like image flip- ping, random rotations, blur, contrast, and bright- ness adjustments. Our findings reveal that certain augmentations, particularly random rotations of up to 15 degrees, vertical image flipping and light blurring, significantly improve the model’s gener- alization to another data set, reducing boundary er- rors and enhancing segmentation accuracy. These results underscore the potential of targeted data augmentations in developing robust, generalizable models for hip joint component segmentation.

With the fast integration of Machine Learning(ML) into several industries, the motivation to develop effective pedagogical strategies for teaching this complex and evolving field has become critical. Machine Learning, once mainly a topic in Computer Science Bachelor programs, is now widely integrated into various majors and introduced at earlier educational stages, including high school and secondary school. However, the reduced research focus on ML pedagogy results in a lack of standard teaching methods compared to other science-related subjects, which have established norms for topic introduction, teaching tools, and assessment methods. With inspiration from other fields, this research aims to look into using interactive visualizations in teaching ML topics, more specifically in teaching Gradient Descent and Principal Component Analysis (PCA). The research includes the development and evaluation of Jupyter Notebooks for introducing these visualizations to students. The targeted student population is composed of Computer Science and Engineering Bachelor students who have not yet followed any Machine Learning courses but have the necessary background knowledge, namely calculus, linear algebra, and statistics knowledge. The evaluation of this teaching method measures the knowledge gained and the motivation of the students, compared to a static version of the materials. The results of the evaluation have shown a significant effect of interactive visualizations on knowledge gained related to PCA. The evaluation has not identified any difference in knowledge gain for gradient descent and learning motivation for both topics. With these results, we contribute to the body of evidence-based teaching methods in Machine Learning and identify further research needed to generalize the effect of interactive visualizations as a teaching method for teaching ML basic concepts.

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.

Osteoarthritis (OA) is a chronic musculoskeletal joint disease that leads to disability. Osteophytes are a hallmark of OA in the knee, characterized by the formation of bone spurs that contribute to joint pain and reduced mobility. This study explores the application of deep learning (DL) techniques for the automatic detection and grading of osteophytes on magnetic resonance (MR) images of the knee. Leveraging the DenseNet-121 and ResNet-50 DL architectures from the Medical Open Network for Artificial Intelligence (MONAI) framework and a dataset, containing 1782 double echo steady-state (DESS) MR images from the Osteoarthritis Initiative (OAI), the study aims to enhance diagnostic accuracy and efficiency in medical imaging analysis. The dataset was split 8:2 for training and validation purposes, respectively. Through a series of numerical experiments, the research evaluates binary classification, region of interest (ROI)-based detection, and multi-class classification models, demonstrating that DenseNet-121 generally outperforms ResNet-50. The five-fold cross-validated binary DenseNet-121 model achieved an area under the receiver operating characteristic curve (ROC AUC) score of 0.90 and a balanced accuracy of 0.82, with a 95% confidence interval (CI) of 0.81-0.83 trained on resampled whole knee images. Moreover, the cross-validated ROI detection models for the patella inferior, superior, and tibia lateral subregions achieved balanced accuracy scores of 0.89 (0.88-0.90 CI), 0.86 (0.85-0.87 CI), and 0.85 (0.84-0.86 CI), respectively. However, the multi-class DenseNet-121 model achieved lower performance, with a balanced accuracy of 0.73 (0.71-0.75 CI), indicating the complexity of multi-class classification in this context. Furthermore, this research did not include hyperparameter optimization, as many settings were kept at their default values, suggesting the possibility for improved results. The cross-validated models were evaluated on an external test set, obtained from the Erasmus Medical Centre, comprising FSPGR-FS images from a significantly younger patient cohort, with a notable class imbalance. The models’ performance on this dataset was significantly lower than their validation results, underscoring the limitations in generalizing to different age demographics and class distributions. External testing underscores the need for more robust models to maintain high performance across diverse datasets and clinical settings. Key contributions of this study include the use of weighted categorical cross-entropy (WCCE) loss functions and analysis of the knee’s subregions to improve detection accuracy. The findings establish a solid foundation for further research, suggesting future work should focus on advanced optimization techniques, mixed imaging sequences in the training dataset, and comparative studies with other established models within the computer vision sector.

In problems with missing data, the data are often considered to be missing at random. This assumption can not be checked from the data. We need to assess the sensitivity of study conclusions to violations of non-identifiable assumptions. This thesis performs Bayesian sensitivity analysis for a missing data model with life time outcomes and covariate information. The outcome distribution is modelled through a Cox model, with a beta process prior on the cumulative hazard function. We run experiments in a simulation study to test the performance of the model in scenarios with simulated data of several sample sizes. We show the validity of the model in the context of Bayesian sensitivity analysis, and propose extensions.

Learning curves in machine learning are graphical representations that depict the relationship between a model's performance and the amount of training data it has been exposed to. They play a fundamental role in obtaining the knowledge and skills across a range of domains. Although there are already quite some researches studying machine learning curves, explaining the importance and practical application of learning curves, we still know very little about the factors that influence the parameters of the learning curve. The aim of this research is to give a better understanding of different factors affecting the parameters of the learning curve. Specifically, we are interested in how the dimensionality of a dataset can influence the parameters of the learning curve. Since learning curves are useful and have several applications, such as estimation of the time required to complete production runs, we would like to know if the dimensionality has any effect on the shapes of learning curves. To conduct the research I applied principal component analysis (PCA) three times with different amount of information preserved to reduce number of dimensions on several datasets and analysed the changes in the parameters of the obtained learning curves. The research showed that potentially there might be some relation between dimensionality and shape of the curve, but only in cases of specific machine learning model. The amount of experiments conducted is not sufficient to make solid conclusions and it is advised to continue with proposed experimental setup, but train machine learning models on increased number of datasets.

Concept drift is an unforeseeable change in the underlying data distribution of streaming data, and because of such a change, deployed classifiers over that data show a drop in accuracy. Concept drift detectors are algorithms capable of detecting such a drift, and unsupervised ones detect drift without needing the data’s actual labels, which can be expensive to obtain. This work is concerned with the implementation and evaluation of two existing unsupervised concept drift detectors based on clustering, UCDD and MSSW, by evaluation on both synthetic and real-world data. Our biggest contribution is in making implementations publicly available. By evaluation, we also realise that UCDD detects drift earlier for simple numerical synthetic datasets, MSSW detects drift earlier for more complex synthetic datasets with categorical features, and none seems suitable for real-world datasets.

Various techniques have been studied to handle unexpected changes in data streams, a phenomenon called concept drift. When the incoming data is not labeled and the labels are also not obtainable with a reasonable effort, detecting these drifts becomes less trivial. This study evaluates how well two data distribution based label-independent drift detection methods, SyncStream and Statistical Change Detection for Multi-Dimensional Data, detect concept drift. This is done by implementing the algorithms and evaluating them side by side on both synthetic and real-world datasets. The metrics used for synthetic datasets are False Positive Rate and Latency; for real-world datasets, Accuracy is used instead of Latency. The experiments show that both drift detectors perform significantly worse on real-world than on synthetic data.

When deployed in production, machine learning models sometimes lose accuracy over time due to a change in the distribution of the incoming data, which results in the model not reflecting reality any longer. A concept drift is this loss of accuracy over time. Drift detectors are algorithms used to detect such drifts. Drift detectors are important as they allow us to detect when a classification model becomes inaccurate. Some possible uses of drift detectors can even go as far as detecting adversarial attacks on machine learning algorithms. The detectors discussed in this paper are Margin Density drift detectors. Their evaluation is made within an unsupervised context, where we assume no testing labels are available. In real world applications of machine learning models, this might often be the case, as finding labels is costly. Experiments in this paper have found that margin density detectors can be useful tools in detecting the first drift for synthetic data, even though parameter tuning must be done to achieve high accuracy for some datasets. In an unsupervised environment with more than one drift, the drift detectors are unreliable as was seen in experiments involving real world data. With this paper comes an implementation of margin density detectors.

Nowadays, software is an integral part of many companies. However, the codebase can grow large and complicated and is often insufficiently documented. To gain insight, tools have been made to infer state machines and process models from software logs. These tools produce different types of models such as automata and Petri nets. The main objective of this research is to determine which tool is the optimal choice for inferring a readable and correct model within reasonable time. Currently, Petri nets and automata are not compared to each other and not all key performance indicators are applicable to both model types. To compare these different concepts, suitable metrics must be identified.

For this work, 8 configurations of 4 programs will be compared. Finite State Machines (FSMs) will be inferred with FlexFringe (AIC), MINT and PRINS (using MINT internally). Petri nets will be mined with ProM using the Inductive Miner, Inductive Miner Infrequent - All Operators, Hybrid-ILP and the Directly Follows miner. The configurations will use 5-folds cross validation to infer models on 9 software logs. Negative traces were synthesised as they were not available. The quality of the models will be measured though inference time, complexity, F2-score, balanced accuracy, fitness and perplexity.

Some of the used metrics were adequate, but others were not suitable. Inference time, F2-score, balanced accuracy could be measured for both FSMs and Petri nets. The complexity was measured with the Petri net $eCFC$ metric and the Cyclomatic complexity CC. The eCFC does not properly express complexity on FSMs. Furthermore, Petri nets can model parallelism, which introduces extra complexity compared to an FSM. This was not adequately expressed by either of these metrics. To measure fitness, both token-based replay and alignment fitness were used. FSMs were converted to Petri nets. Token-based replay (TBR) fitness was not an expressive metric for the FSMs, as the concept of tokens did not carry over well. In addition to this, the external implementation of TBR fitness was flawed for the specific structure of the converted FSMs. Alignment-based fitness is the superior fitness metric as it does not rely on the notion of tokens, which the FSMs do not have. Unfortunately, the time and memory needed for alignment computations was too large for some models. Lastly, the perplexity FSM metric was successfully adapted for Petri nets. It expresses the difference in structure and could be tailored even further for the purposes of comparison by adjusting its parameters.

The results of the comparison showed that almost all configurations could complete inference in feasible time and memory. The time out was set at 4 hours and the available memory was \texttt{16GB}. The MINT and PRINS ran out of memory on one of the larger logs and timed out for one other set. Hybrid-ILP timed out for 2 sets. All other configurations completed inference for all data sets in under 40 seconds. PRINS and MINT boasted excellent performance across all correctness metrics, and were only outperformed by FlexFringe and the Directly Follows miner on perplexity. PRINS and MINT were most suitable for modelling traces of data sets with a low trace similarity and generalising to identify new traces. However, MINT models were a lot larger than those of any other configurations and PRINS models are generally many times larger than MINT models. So, if complexity of models is a big concern, FlexFringe and the Directly Follows miner offer the smaller models, at the cost of a small amount of performance for most sets. However, these two tools perform poorly on sets that are both incomplete and contain dissimilar traces. If time is of the essence, one should use FlexFringe, the Directly Follows miner of one of the Inductive Miners. The Petri net miners used, were not designed to introduce new behaviour in a model. Therefore, FlexFringe is preferable for modelling an incomplete log.

Flood simulations can give insight into the consequences of flood scenario's and can help to create hazard- and risk maps to support decision-making in flood risk management and in crisis management. 2D hydrodynamic simulations give accurate descriptions of the propagation of a flood and rely on advanced numerical methods to solve a set of physics-based mathematical equations. A drawback of these models is that they can be computationally expensive with run times in the order of hours or days depending on the time and spatial resolutions. In this study we explore the use of deep learning techniques in a surrogate model for 2D flood simulation. We propose and test a deep learning-based surrogate modelling framework that can be used to train a deep learning-based surrogate model. Once trained, the surrogate model can be used as a substitute for the hydrodynamic model with the advantage of being much more efficient in terms of run time and can be of great value in for example crisis situations. For training, a data set of expensive 2D hydrodynamic simulations was created using the SOBEK software program. Such simulations require a lot of input data, such as input parameter maps specifying the terrain over the computational grid and boundary conditions. To make training data-efficient, a sampling strategy was used for the input of the flood simulations. Three deep learning architectures were trained and tested. The first two architectures are feed-forward networks and the third architecture is of recurrent network type. These networks contain convolutional neural network (CNN) architectures with an encoder-decoder structure to make patch-level predictions of the flood characteristics in time. These patches contain a small section of the flood prone area and an encoder network is used to extracts coarse feature maps from this data that is then refined by a decoder network to create a prediction of the flood propagation. Using patches has the advantage of making a surrogate model able to create flood simulations over prone areas without restrictions on size or shape by tiling the output patches with flow predictions. Also it allows the surrogate to focus only on regions where the flood has reached and not on the regions where no water has arrived. It was found that with the recurrent architecture, the surrogate model was most capable of emulating the ground truth flood simulations in the test simulation. This trained network architecture was used in a case study where the surrogate was applied to create flood simulations in a small dike ring in the Netherlands. This shows that the surrogate modelling framework can be used to train a deep learning-based surrogate and, once trained, can be used to create flood simulations similar to hydrodynamic simulations. However, two main challenges were identified in using such data-driven deep learning-based surrogates. Firstly, keeping the predictions of the flood characteristics accurate enough to avoid large error propagation. Secondly, accurately generating large amounts of data from relatively little information present in the boundary condition and terrain.

Year after year, the amount of network intrusions and costs associated to them rises. Research in this area is, therefore, of high importance and provides valuable insight in how to prevent or counteract intrusions. Machine learning algorithms seem to be a promising answer for automated network intrusion detection, as their results often reach upwards of 99\%+ on datasets. Yet even with these results, the problem does not seem to be solved as the same models do not reach similar scores on real live network traffic. This indicates a problem with the datasets. In this work, we explored six recent network intrusion datasets and measured the quality of them from a design perspective and through a practical binary classification approach. Furthermore, we explored the need for complex classification models on these datasets, as research has shifted more towards using black-box models as opposed to white-box models. Through a literature study on the quality metrics of a dataset, we found there is a general lack of agreement amongst researchers regarding what makes a dataset good and realistic. We also found areas in which datasets are often lacking in and provide concrete advice on how to upgrade the quality of these datasets. For our practical classification approach, we built a general classification pipeline using a Random Forest. Three feature sets were tested, two of which form an ablation study to measure the effects of trigrams on the classification score. On all datasets the general classification model reaches good classification results (>90\%+) and for three datasets even reached state-of-the-art results. The ablation study yielded a positive effect of using trigrams for classification on the datasets. Our white-box approach performed on par or better than most black-box techniques. We conclude that black-box models are unnecessary in this problem context and that the techniques should shift back to white-box approaches. Next, we attempted to link the quality of dataset methodologies to the difficulty of obtaining state-of-the-art classification results. Apart from the complexity of the attack vector and benign traffic variety, we found no further properties that define this relationship. Finally, this work started with the assumption that real live network traffic is more complex and therefore more difficult to classify well on than most available datasets. Newer datasets do show improvements over older datasets and our classification results corroborated the validity of the assumption.

The field of finance is an interesting field in which much research takes place. In particular, its sub-field of modeling the dynamics of order books is an interesting field, since it translates into modeling the behaviour of traders on the market. Most of the models proposed in this field can be divided into stochastic models or machine learning models. The problem with the former ones is that they often make assumptions that do not propagate to practical applications, while the problem with the latter is that the models are often incomprehensible, even though they yield good results. This work tries to combine both of those fields to maintain the transparency of stochastic models and the accuracy of machine learning models. We target the limit order book, which is a book used in financial markets to place orders for financial securities in. It is a challenge to both generate realistic order books and predict them as well, since the model should be able to understand the underlying dynamics of order books to do this. We present a deep-learning-based model that is able to generate order books, through the use of a generative adversarial network (GAN) like architecture, and is able to predict order books using a similar architecture for the generator. Our models are able to outperform traditional generative and predictive models in both instances, while it remains transparent in its calculation: it is still able to extract useful and interpretable features from the underlying data set. These features, like the probabilities for up or down movements, or stagnations, or the rates at which the variables within the order book move, can be used by traders or algorithms to make sense out of the predictions. Traders can, for example, ask themselves whether the extracted features are representative of the analyzed asset and can, through this question, build additional confidence in the model. Another use case could be to utilize those features as inputs to other models. To our knowledge, this is the first work that proposes a combination of a stochastic and machine learning model to learn the dynamics of the order book.

Mutual predictability shows itself as a contributing factor to mutual trust and is known to improve the effectiveness in a human-agent teamwork setting. As team members communicate to coordinate the team through the task, the question arises as to what information the human should share to be predictable to an agent. To experiment with measuring the predictability, defined as to what extent the agent can anticipate the actions of the human, we used the Blocks World for Teams (BW4T) task. The two information types shared are intentions and world knowledge. The predictability is evaluated in the background by the agent logging, with automatic help from the human agent, a sequence containing the human performed actions. The agent can assign 2 probabilities to a human performed action. The first indicates with what probability the agent could say the human chose this action. The second indicates with what probability the agent could say this action led to this outcome. These probabilities then vary based on whether the information sent beforehand implied this action. The experiment has 4 different cases: sharing no information, sharing intentions, sharing world knowledge, and sharing both types. It is shown that sharing intentions contributes the most to higher predictability, due to these messages being the most effective at implying the future action of the human. World knowledge and sharing both types have less effect on predictability. We speculate that this is because of the larger amount of messages to share as well as when to share them, which overloads the human.

Audio fingerprinting is a technique that allows for fast identification of music. Research concerning this technique first emerged around the 2000s and has lead to several applications, like Shazam. More recently, developments in this area have slowed down, even though there are still new challenges emerging. This paper investigates one of these challenges, music identification of movies, in a systematic way using the open-source fingerprinting framework called Panako. First, clips containing music were extracted from movies and queried using the default settings for Panako. Then, movie soundtracks were modified by layering noise over them, or by time-stretching and pitch-shifting them, and the performance of Panako on these modified audio signals was evaluated using a benchmark. Finally, the best configurations from the previous step were again queried using actual movie clips. These tests showed that both the default configuration and the configurations that performed best on the synthesised data perform poorly in movie music identification. Less than 10\% of the clips were identified correctly. The limited scope of this research, combined with the results gathered, show that there should be further investigation into the suitability of Panako for movie music identification.

This paper presents the findings of a benchmark performed on the audio fingerprinting framework OLAF in the context of movie music. The goal is to find a music identification framework suitable for automatically identifying a song from a movie clip. This research aims to find how well OLAF performs in this context with regard to the criteria determined in the benchmark and to see if the performance can be increased. The OLAF framework makes use of parameters, which are individually tweaked and benchmarked with synthesised data. These findings are then combined to find a set of parameters that is used to perform this same benchmark on real clips from movies. The found parameter setup slightly increased the performance, going from 1 true positive and 133 false negatives in the original setup to 5 true positives, 3 false positives and 126 false negatives.

Music indexing, the practice of identifying songs contained in an audio sample, is an approach that is widely used. As an underlying technique, "audio fingerprinting" can be used. In this technique, an audio sample is converted to a fingerprint; a smaller representation of the audio. This fingerprint is compared to a database of fingerprinted songs in order to retrieve the original song. In this research, we aim to answer the question of how music in movies can be identified using an audio fingerprinting platform called Audioneex. Two configurable parameters in the framework have been varied, of which the configurations have been evaluated in terms of a benchmark that has been established on synthesized data. The best performing configuration found in the limits of the parameters has been compared to the base configuration of Audioneex. Based on the results, the selected configuration improved the performance of the Audioneex framework by 1 match for one matching algorithm and by 3 matches for the other available matching algorithm.

Audio fingerprinting is one of the standard solutions for music identification. The underlying technique is designed to be robust to signal degradation such that music can be identified despite its presence. One of the newly emerged applications of a possibly challenging nature is music identification in movies. This paper examines the audio fingerprinting framework Dejavu by evaluating its performance against an existing benchmark created for the context of music identification in movies. The results show that Dejavu’s performance matches the expectations derived from the implementation and previous testing, and can be reconfigured to improve the performance in terms of the benchmark.

Background: The covid-19 pandemic has overwhelmed hospitals worldwide and clinical prediction models may assist in timely identification of covid-19 patients at risk for clinical deterioration, i.e. `early warning'. In this article, we report on the development and validation of a new early warning model that predicts unplanned ICU admission or unexpected death within 24 hours from the moment of prediction, specifically for covid-19 patients. We compared the performance with two well-known and widely used early warning scores (EWSs), i.e. the Modified Early Warning Score (MEWS) [2] and National Early Warning Score (NEWS) [3]. Methods: We collected electronic medical record (EMR) data from covid-19 patients admitted to six Dutch hospitals between February 2020 and May 2021. We defined the clinical endpoint as a surrogate of unplanned ICU admission or unexpected death. To examine the added value of including non-linear predictor-outcome relations, we trained both a (linear) logistic regression (LR) and a (non-linear) random forest (RF) model. We included predictors based on patient demographics, vital signs and laboratory test results. We validated the models retrospectively in a `leave-one-hospital-out' cross-validation (LOHO-CV) procedure. Furthermore, we simulated a prospective validation by splitting all included patients admitted before and after August 1 2020 and simulated as if the models would have been developed based on the data collected until August 2020 and implemented during the remaining study period. Additionally, we examined different strategies for monthly model updating. We evaluated model discrimination and calibration for the proposed models as well as the traditional EWSs, and performed a decision curve analysis [22]. Importance of individual predictors was quantified using SHAP values [13]. Findings: In the retrospective validation, the LR model yielded a significant improvement in partial area under the receiver operating curve (pAUC) compared to the traditional EWSs in four of the six included hospitals, and in all hospitals by the RF model. In the simulated prospective validation, significant improve- ment was shown in two and four hospitals by the LR and RF models, respectively. Without any model updating, both model showed risk overestimation. We proposed a combination of monthly model retraining and hospital-specific re-calibration that could correct for this miscalibration effectively. In the decision curve analysis, the proposed models outperformed the traditional EWS in terms of net benefit (NB) over a wide range of clinically relevant model thresholds. Interpretation We have derived and validated a new early warning model specifically for covid-19 patients that outperformed traditional EWSs and showed good generalizability over different Duch hospitals. Also, we introduced SpO2-to-O2 ratio as an important marker for disease severity in covid-19 patients. Finally, we showed the importance of repeated model updating when developing medical prediction models in the midst of the covid-19 pandemic and proposed an effective model updating strategy. Future research is needed to validate the model outside the Netherlands. Background: The covid-19 pandemic has overwhelmed intensive care units (ICUs) worldwide. Improved prediction of a covid-19 patient's risk of dying may assist decision making in the intensive care unit (ICU) setting. In contrast to traditional mortality models like APACHE II [1] and SAPS II [2], dynamic mortality models allow for repeated risk stratication of patients throughout the ICU stay. Earlier works [3, 4, 5] in dynamic mortality modelling show promising results, although most of these works propose models for the general (non-covid) ICU population and use relatively long or unspecied prediction horizons. In this study, we report on the development and retrospective validation of a model for dynamic, near-term mortality for critically ill covid-19 patients. Methods: We collected EMR data from 3 481 ICU admissions with a covid-19 infection from the Dutch Data Warehouse (DDW) [6], coming from 25 dierent ICUs in the Netherlands. We extracted daily samples of each patient and trained both a linear (logistic regression) and non-linear (random forest) model to predict in-ICU mortality within 24 hours from the moment of prediction. Isotonic regression was used to re-calibrate the predictions of the trained models. We evaluated the models in a leave-one-ICU-out (LOIO) cross-validation procedure. Findings Validation in 21 out of 25 and 18 out of 25 ICUs yielded an area under the receiver operating characteristic curve (AUROC) >0.80 for the logistic regression and random forest model, respectively. In the four hospitals that yielded an AUROC<0.8, local differences in protocols concerning discontinuation of treatment may have played a role. The re-calibrated model estimations showed good calibration for both models (calibration intercept = −0.12, slope = 0.87 for logistic regression and intercept = −0.05, slope = 0.82 for random forest). Interpretation: This study is different from previous works on dynamic mortality prediction in the ICU as we presented a model speciffcally for covid-19 patients and introduced near-term mortality predictions (compared to long-term or in-hospital mortality). The predictions were calculated based on a mixture of static information (e.g. age and sex) and dynamic information (e.g., vital signs and laboratory values) and the importance of individual predictors was quantied using SHAP values [7], where we found FiO2, oxygen saturation and pH to be important predictors. The potential clinical utility of dynamic mortality models such as guidance in resource allocation and real-time patient benchmarking could be topics for future research.

Recommender Systems assist the user by suggesting items to be consumed based on the user's history. The topic of diversity in recommendation gained momentum in recent years as additional criterion besides recommendation accuracy, to improve user satisfaction. Accuracy and diversity in recommender systems coexist in a delicate trade-off due to the complexity in capturing user tastes through a limited amount of interactions. Graphs have been employed for recommendation, given their ability to efficiently represent user-item interactions. Graph convolutions, as learning over graphs tools, have reached state-of-the-art accuracy on recommender system benchmarks. However, the potential of graph convolutions to improve the accuracy-diversity trade-off is unexplored. Here, we develop a model that learns from a nearest neighbor and a furthest neighbor graph via a joint convolutional model to establish a novel accuracy-diversity trade-off for recommender systems. In detail, the nearest neighbor graph connects entities (users or items) based on their similarities and is responsible for improving accuracy, while the furthest neighbor graph connects entities based on their dissimilarities and is responsible for diversifying recommendations. The information between the two convolutional modules is balanced already in the training phase through a regularizer inspired by multi-kernel learning. Numerical experiments on three benchmark datasets showed the joint convolutional model can improve substantially the catalog coverage or the diversity among recommended items; or boost both by a lesser amount. We compared our model against state-of-the-art accuracy-oriented algorithms, showing diversity gains up to seven times by trading as little as 1\% in accuracy. We also compared the joint model against algorithms proposing a different accuracy-diversity trade-off, evidencing our model achieves better accuracy while preserving a wide diversity range. Our findings highlight that the joint convolutional model offers a balance in each setting that is difficult to be achieved with a single model.