Abstract Extracting and identifying latent topics in large text corpora have gained increasing importance Natural Language Processing (NLP). Most models, whether probabilistic models similar to Latent Dirichlet Allocation (LDA) or neural topic follow the same underlying approach of interpretability extraction. We propose a method that incorporates deeper understanding both sentence document themes, goes beyond simply analyzing word frequencies data. Through simple corpus expansion, our model...

Tinnitus is the subjective perception of a sound in absence corresponding external acoustic stimuli. Research highlights influence sensorimotor system on tinnitus perception. Associated neuronal processes, however, are insufficiently understood and it remains unclear how at which hierarchical level interacts with tinnitus-processing auditory system. We therefore asked 23 patients suffering from chronic (11 males) to perform specific exercises, aimed relaxing or tensing jaw area, temporarily...

Abstract Tuning of model-based boosting algorithms relies mainly on the number iterations, while step-length is fixed at a predefined value. For complex models with several predictors such as Generalized additive for location, scale and shape (GAMLSS), imbalanced updates predictors, where some distribution parameters are updated more frequently than others, can be problem that prevents submodels to appropriately fitted within limited iterations. We propose an approach using adaptive (ASL)...

Boosting techniques from the field of statistical learning have grown to be a popular tool for estimating and selecting predictor effects in various regression models can roughly separated two general approaches, namely gradient boosting likelihood-based boosting. An extensive framework has been proposed order fit generalized mixed based on boosting, however case cluster-constant covariates approaches tend mischoose variables selection step leading wrong estimates. We propose an improved...

Joint models are a powerful class of statistical which apply to any data where event times recorded alongside longitudinal outcome by connecting and time-to-event within joint likelihood allowing for quantification the association between two outcomes without possible bias. In order make feasible regularization variable selection, boosting algorithm has been proposed, fits using component-wise gradient techniques. However, these methods have well-known limitations, i.e., they provide no...

Abstract Mixture Density Networks (MDN) belong to a class of models that can be applied data which cannot sufficiently described by single distribution since it originates from different components the main unit and therefore needs mixture densities. In some situations, however, MDNs seem have problems with proper identification latent components. While these issues extent contained using custom initialization strategies for network weights, this solution is still less than ideal involves...

Component-wise gradient boosting algorithms are popular for their intrinsic variable selection and implicit regularization, which can be especially beneficial very flexible model classes. When estimating generalized additive models location, scale shape (GAMLSS) by means of a component-wise algorithm, an important part the estimation procedure is to determine relative complexity submodels corresponding different distribution parameters. Existing methods either suffer from computationally...

<title>Abstract</title> Generalised additive mixed models are a common tool for modelling of grouped or longitudinal data where random effects incorporated into the model in order to account within-group inter-individual correlations. As an alternative established penalised maximum likelihood approaches, several different types boosting routines have been developed make more demanding situations manageable. However, when estimating with component-wise gradient boosting, and fixed compete...

Modeling longitudinal data (e.g., biomarkers) and the risk for events separately leads to a loss of information bias, even though underlying processes are related each other. Hence, popularity joint models time-to-event-data has grown rapidly in last few decades. However, it is quite practical challenge specify which part model single covariates should be assigned as this decision usually made based on background knowledge. In work, we combined recent developments from field gradient...

Abstract Joint models for longitudinal and time-to-event data simultaneously model information to avoid bias by combining usually a linear mixed with proportional hazards model. This class has seen many developments in recent years, yet joint including spatial predictor are still rare the traditional formulation of part is accompanied computational challenges. We propose piecewise exponential hazard using counting process representation structured additive predictors able estimate...

Bayesian structured additive quantile regression is an established tool for regressing outcomes with unknown distributions on a set of explanatory variables and/or when interest lies effects the more extreme values outcome. Even though variable selection exists, its scope limited. We propose use Normal Beta Prime Spike and Slab (NBPSS) prior in to aid researcher not only but also effect selection. compare NBPSS approach statistical boosting regression, current standard automated simulation...

Selection of relevant fixed and random effects without prior choices made from possibly insufficient theory is important in mixed models. Inference with current boosting techniques suffers biased estimates the inflexibility selection. This paper proposes a new inference method "BayesBoost" that integrates Bayesian learner into gradient simultaneous estimation selection linear The introduces novel strategy for effects, which allows computationally fast slopes even high-dimensional data...

Joint models for longitudinal and time-to-event data have seen many developments in recent years. Though spatial joint are still rare the traditional proportional hazards formulation of part model is accompanied by computational challenges. We propose a with piece-wise exponential hazard using counting process representation structured additive predictors able to estimate (non-)linear, random effects. Its capabilities assessed simulation study comparing our approach an established one...

Model-based component-wise gradient boosting is a popular tool for data-driven variable selection. In order to improve its prediction and selection qualities even further, several modifications of the original algorithm have been developed, that mainly focus on different stopping criteria, leaving actual mechanism untouched. We investigate prediction-based mechanisms step in model-based boosting. These approaches include Akaikes Information Criterion (AIC) as well rule relying test error...

Extracting and identifying latent topics in large text corpora has gained increasing importance Natural Language Processing (NLP). Most models, whether probabilistic models similar to Latent Dirichlet Allocation (LDA) or neural topic follow the same underlying approach of interpretability extraction. We propose a method that incorporates deeper understanding both sentence document themes, goes beyond simply analyzing word frequencies data. This allows our model detect may include uncommon...

Abstract Tinnitus is the subjective perception of a sound in absence corresponding external acoustic stimuli. Research highlights influence sensorimotor system on tinnitus perception. Associated neuronal processes, however, are insufficiently understood and it remains unclear how at which hierarchical level interacts with tinnitus-processing auditory system. We therefore asked 23 patients suffering from chronic (11 males) to perform specific exercises, aimed relaxing or tensing jaw area,...

Model-based component-wise gradient boosting is a popular tool for data-driven variable selection. In order to improve its prediction and selection qualities even further, several modifications of the original algorithm have been developed, that mainly focus on different stopping criteria, leaving actual mechanism untouched. We investigate prediction-based mechanisms step in model-based boosting. These approaches include Akaikes Information Criterion (AIC) as well rule relying test error...