technicalreports.bib

@article{seiler2024,
  author = {Seiler, Johannes and M{\"u}ller, Benjamin and G{\"u}nther, Isabel and Wetscher, Mattias and Stauffer, Reto and Umlauf, Nikolaus and Harttgen, Kenneth},
  title = {Co-occurrence Patterns of Anemia, Acute, and Chronic Malnutrition in sub-{S}aharan {A}frica},
  year = {2024},
  month = {10},
  journal = {Communications Medicine},
  note = {Status: submitted}
}

@techreport{zackl2024,
  author = {Zackl, Constantin and Zopoglou, Maria and Stauffer, Reto and Ausserhofer, Markus and Ijetoselsteijn, Marieke and Sturm, Gregor and de Miranda, Noel and Finotello, Francesca},
  year = {2024},
  month = {09},
  pages = {1--41},
  title = {{spacedeconv}: deconvolution of tissue architecture from spatial transcriptomics},
  institution = {Research Square},
  journal = {Genome Biology},
  abstract = {Investigating tissue architecture is key to understanding tissue function in health and disease. While spatial omics technologies enable the study of cell transcriptomes within their native context, they often lack single-cell resolution. Deconvolution methods can computationally infer tissue composition from spatial transcriptomics data, but differences in their workflows complicate their use and comparison. We developed spacedeconv, a unified interface to different deconvolution methods that additionally supports data preprocessing, visualization, and analysis of cell communication and multimodal data. Here, we demonstrate how spacedeconv streamlines the investigation of the cellular and molecular underpinnings of tissue architecture in different organisms and tissue contexts.},
  doi = {10.21203/rs.3.rs-5102166/v1},
  status = {submitted}
}

@techreport{wetscher2024,
  title = {Stagewise Boosting Distributional Regression},
  author = {Mattias Wetscher and Johannes Seiler and Reto Stauffer and Nikolaus Umlauf},
  year = {2024},
  month = {5},
  number = {2405.18288},
  institution = {arXiv.org E-Print Archive},
  type = {arXiv},
  doi = {10.48550/arXiv.2405.18288},
  abstract = {Forward stagewise regression is a simple algorithm that can be used to estimate regularized models. The updating rule adds a small constant to a regression coefficient in each iteration, such that the underlying optimization problem is solved slowly with small improvements. This is similar to gradient boosting, with the essential difference that the step size is determined by the product of the gradient and a step length parameter in the latter algorithm. One often overlooked challenge in gradient boosting for distributional regression is the issue of a vanishing small gradient, which practically halts the algorithm's progress. We show that gradient boosting in this case oftentimes results in suboptimal models, especially for complex problems certain distributional parameters are never updated due to the vanishing gradient. Therefore, we propose a stagewise boosting-type algorithm for distributional regression, combining stagewise regression ideas with gradient boosting. Additionally, we extend it with a novel regularization method, correlation filtering, to provide additional stability when the problem involves a large number of covariates. Furthermore, the algorithm includes best-subset selection for parameters and can be applied to big data problems by leveraging stochastic approximations of the updating steps. Besides the advantage of processing large datasets, the stochastic nature of the approximations can lead to better results, especially for complex distributions, by reducing the risk of being trapped in a local optimum. The performance of our proposed stagewise boosting distributional regression approach is investigated in an extensive simulation study and by estimating a full probabilistic model for lightning counts with data of more than 9.1 million observations and 672 covariates.}
}

@techreport{stauffer2019,
  author = {Reto Stauffer and Matthias Dusch and Fabien Maussion and Georg J. Mayr},
  title = {Making Surface Wind Diagnostics Reproducible, Comparable and Scalable With {foehnix}},
  year = 2019,
  month = 10,
  journal = {Bulletin of the American Meteorological Society},
  institution = {American Meteorological Society},
  abstract = {Given wind measurements and possibly additional information: how can one identify different wind regimes at least as well as human experts but with reproducibility, scalability and a degree of certainty assigned to the classification? While algorithms from the field of machine learning exist to perfor m such a classification task they have been difficult to use by non-specialists. The open-source software package foehnix available in two popular languages - Python and R - gives them tools to perform the whole classification workflow with ease in a few lines of code.  Extensive documentation and working examples add to the ease of use. The paper showcases oehnix identifying strong gusty winds that descend behind a mountain ran ge in California and can catastrophically fan wildfires. Such winds occur globally downstream of mountain s and straits and change temperature, humidity and air quality. They modify the local climate, affect agr iculture e.g. by killing rice plants, influence pollutants/aerosol mixing ratios, and provide ideal spots at which to generate electricity with wind farms or develop kitesurfing or windsurfing infrastructure. foehnix can also classify other wind regimes such as see breezes and valley winds and flexibly inc orporate available information. Applying it makes results from different regions comparable and handling of extensive data sets possible for users from many different fields.}
}