arXiv subjects

Stefanie Schwaar

Publications and source records attributed to Stefanie Schwaar.

7 recordsLinked to original sources

arXiv2025

Study on Text Classification for Public Administration

In German public administration, there are 45 different offices to which incoming messages need to be distributed. Since these messages are often unstructured, the system has to be based at least partly on message content. For public service no data are given so far and no pretrained model is available. The data we used are conducted by Governikus KG and are of highly different length. To handle those data with standard methods different approaches are known, like normalization or segmentation. However, text classification is highly dependent on the data structure, a study for public administration data is missing at the moment. We conducted such a study analyzing different techniques of classification based on segments, normalization and feature selection. Thereby, we used different methods, this means neural nets, random forest, logistic regression, SVM classifier and SVAE. The comparison shows for the given public service data a classification accuracy of above 80\% can be reached based on cross validation. We further show that normalization is preferable, while the difference to the segmentation approach depends mainly on the choice of algorithm.

stat.AP↗

arXiv2025

Estimation of Change Points for Non-linear (auto-)regressive processes using Neural Network Functions

In this paper, we propose a new test for the detection of a change in a non-linear (auto-)regressive time series as well as a corresponding estimator for the unknown time point of the change. To this end, we consider an at-most-one-change model and approximate the unknown (auto-)regression function by a neuronal network with one hidden layer. It is shown that the test has asymptotic power one for a wide range of alternatives not restricted to changes in the mean of the time series. Furthermore, we prove that the corresponding estimator converges to the true change point with the optimal rate OP (1/n) and derive the asymptotic distribution. Some simulations illustrate the behavior of the estimator with a special focus on the misspecified case, where the true regression function is not given by a neuronal network. Finally, we apply the estimator to some financial data.

math.ST↗

arXiv2021

A Generalised Linear Model Framework for $β$-Variational Autoencoders based on Exponential Dispersion Families

Although variational autoencoders (VAE) are successfully used to obtain meaningful low-dimensional representations for high-dimensional data, the characterization of critical points of the loss function for general observation models is not fully understood. We introduce a theoretical framework that is based on a connection between $β$-VAE and generalized linear models (GLM). The equality between the activation function of a $β$-VAE and the inverse of the link function of a GLM enables us to provide a systematic generalization of the loss analysis for $β$-VAE based on the assumption that the observation model distribution belongs to an exponential dispersion family (EDF). As a result, we can initialize $β$-VAE nets by maximum likelihood estimates (MLE) that enhance the training performance on both synthetic and real world data sets. As a further consequence, we analytically describe the auto-pruning property inherent in the $β$-VAE objective and reason for posterior collapse.

cs.LG↗

arXiv2020

Adaptive Quantile Computation for Brownian Bridge in Change-Point Analysis

As an example for the fast calculation of distributional parameters of Gaussian processes, we propose a new Monte Carlo algorithm for the computation of quantiles of the supremum norm of weighted Brownian bridges. As it is known, the corresponding distributions arise asymptotically for weighted CUSUM statistics for change-point detection. The new algorithm employs an adaptive (sequential) time discretization for the trajectories of the Brownian bridge. A simulation study shows that the new algorithm by far outperforms the standard approach, which employs a uniform time discretization.

stat.CO↗

arXiv2020

A Data-driven Change-point Estimator

The q-weighted CUSUM and their corresponding estimator are well known statistics for change-point detection and estimation. They have the difficulty that the performance is highly dependent on the location of the change. An adaptive estimator with data-driven weights is presented to overcome this problem, and it is shown that the corresponding adaptive change-point tests are valid.

stat.AP↗

arXiv2020

A lower bound for the ELBO of the Bernoulli Variational Autoencoder

We consider a variational autoencoder (VAE) for binary data. Our main innovations are an interpretable lower bound for its training objective, a modified initialization and architecture of such a VAE that leads to faster training, and a decision support for finding the appropriate dimension of the latent space via using a PCA. Numerical examples illustrate our theoretical result and the performance of the new architecture.

cs.LG↗

arXiv2018

Application of Clustering Methods to Anomaly Detection in Fibrous Media

The paper considers the problem of anomaly detection in 3D images of fibre materials. The spatial Stochastic Expectation Maximisation algorithm and Adaptive Weights Clustering are applied to solve this problem. The initial 3D grey scale image was divided into small cubes subject to clustering. For each cube clustering attributes values were calculated: mean local direction and directional entropy. Clustering is conducted according to the given attributes. The proposed methods are tested on the simulated images and on real fibre materials.

stat.AP↗