arXiv subjects

Steffen Goebbels

Publications and source records attributed to Steffen Goebbels.

5 recordsLinked to original sources

arXiv2025

Column-generation for a two-dimensional multi-criteria bin-packing problem

In this study, we examine a two-dimensional bin-packing problem in printed circuit board manufacturing. Among other objectives, the number of bins, but also the number of different bin layouts, is to be minimized. As the running times of an earlier MIP presentation are only acceptable for small problem instances, we will now discuss a branch-and-price approach by using an adapted Ryan-Foster-branching. The pricing problem computes the layouts, separating the time-consuming constraints from the master problem.

cs.DS↗

arXiv2023

A Simple Bias Reduction for Chatterjee's Correlation

Chatterjee's rank correlation coefficient $\xi_n$ is an empirical index for detecting functional dependencies between two variables $X$ and $Y$. It is an estimator for a theoretical quantity $\xi$ that is zero for independence and one if $Y$ is a measurable function of $X$. Based on an equivalent characterization of sorted numbers, we derive an upper bound for $\xi_n$ and suggest a simple normalization aimed at reducing its bias for small sample size $n$. In Monte Carlo simulations of various models, the normalization reduced the bias in all cases. The mean squared error was reduced, too, for values of $\xi$ greater than about 0.4. Moreover, we observed that non-parametric confidence intervals for $\xi$ based on bootstrapping $\xi_n$ in the usual n-out-of-n way have a coverage probability close to zero. This is remedied by an m-out-of-n bootstrap without replacement in combination with our normalization method.

stat.ME↗

arXiv2021

ReLU activated Multi-Layer Neural Networks trained with Mixed Integer Linear Programs

In this paper, it is demonstrated through a case study that multilayer feedforward neural networks activated by ReLU functions can in principle be trained iteratively with Mixed Integer Linear Programs (MILPs) as follows. Weights are determined with batch learning. Multiple iterations are used per batch of training data. In each iteration, the algorithm starts at the output layer and propagates information back to the first hidden layer to adjust the weights using MILPs or Linear Programs. For each layer, the goal is to minimize the difference between its output and the corresponding target output. The target output of the last (output) layer is equal to the ground truth. The target output of a previous layer is defined as the adjusted input of the following layer. For a given layer, weights are computed by solving a MILP. Then, except for the first hidden layer, the input values are also modified with a MILP to better match the layer outputs to their corresponding target outputs. The method was tested and compared with Tensorflow/Keras (Adam optimizer) using two simple networks on the MNIST dataset containing handwritten digits. Accuracies of the same magnitude as with Tensorflow/Keras were achieved.

cs.LG↗

arXiv2020

On Sharpness of Error Bounds for Multivariate Neural Network Approximation

Single hidden layer feedforward neural networks can represent multivariate functions that are sums of ridge functions. These ridge functions are defined via an activation function and customizable weights. The paper deals with best non-linear approximation by such sums of ridge functions. Error bounds are presented in terms of moduli of smoothness. The main focus, however, is to prove that the bounds are best possible. To this end, counterexamples are constructed with a non-linear, quantitative extension of the uniform boundedness principle. They show sharpness with respect to Lipschitz classes for the logistic activation function and for certain piecewise polynomial activation functions. The paper is based on univariate results in (Goebbels, St.: On sharpness of error bounds for univariate approximation by single hidden layer feedforward neural networks. Results Math 75 (3), 2020, article 109, https://rdcu.be/b5mKH).

math.FA↗

arXiv2020

On Sharpness of Error Bounds for Single Hidden Layer Feedforward Neural Networks

A new non-linear variant of a quantitative extension of the uniform boundedness principle is used to show sharpness of error bounds for univariate approximation by sums of sigmoid and ReLU functions. Single hidden layer feedforward neural networks with one input node perform such operations. Errors of best approximation can be expressed using moduli of smoothness of the function to be approximated (i.e., to be learned). In this context, the quantitative extension of the uniform boundedness principle indeed allows to construct counter examples that show approximation rates to be best possible. Approximation errors do not belong to the little-o class of given bounds. By choosing piecewise linear activation functions, the discussed problem becomes free knot spline approximation. Results of the present paper also hold for non-polynomial (and not piecewise defined) activation functions like inverse tangent. Based on Vapnik-Chervonenkis dimension, first results are shown for the logistic function.

math.FA↗