Searcharxiv⌕ Search

arXiv subjects

Erik Müller

Publications and source records attributed to Erik Müller.

3 recordsLinked to original sources

Adaptive Constraint Sampling for Stochastic Gradient Methods in Constrained Optimization

This paper proposes an adaptive constraint-sampling technique for stochastic gradient-based methods to efficiently solve constrained optimization problems with a large number of inequality constraints. The proposed constraint-sampling technique maintains a set of recently violated constraints and draws the constraint used in each iteration uniformly either from this set or from the set of all constraints. This keeps its per-iteration arithmetic overhead independent of the number of constraints. In simulations on synthetic convex optimization problems and in the synthesis of a control barrier function for the Dubins Car, the constraint-sampling technique is shown to significantly outperform uniform constraint sampling.

math.OC↗

GraphBench: Next-generation graph learning benchmarking

Machine learning on graphs has made substantial progress across domains such as molecular property prediction and chip design. Yet benchmarking practices remain fragmented, often relying on narrow, task-specific datasets and inconsistent evaluation protocols, hindering reproducibility and broader progress. With the recent popularity of graph foundation models, these weaknesses have become apparent, as existing benchmarks are insufficient for thorough evaluation. To address these challenges, we introduce GraphBench, a comprehensive benchmark suite spanning diverse real-world domains and task settings, including node-level, edge-level, graph-level, and generative tasks. GraphBench provides standardized evaluation protocols, including consistent dataset splits and metrics for assessing out-of-distribution generalization across selected tasks, as well as a unified hyperparameter-tuning framework. We further evaluate GraphBench with recent message-passing neural networks and graph transformer models, establishing principled baselines for future research. See www.graphbench.io for further details.

cs.LG↗

Comparison of the performance and reliability between improved sampling strategies for polynomial chaos expansion

As uncertainty and sensitivity analysis of complex models grows ever more important, the difficulty of their timely realizations highlights a need for more efficient numerical operations. Non-intrusive Polynomial Chaos methods are highly efficient and accurate methods of mapping input-output relationships to investigate complex models. There is substantial potential to increase the efficacy of the method regarding the selected sampling scheme. We examine state-of-the-art sampling schemes categorized in space-filling-optimal designs such as Latin Hypercube sampling and L1-optimal sampling and compare their empirical performance against standard random sampling. The analysis was performed in the context of L1 minimization using the least-angle regression algorithm to fit the GPCE regression models. Due to the random nature of the sampling schemes, we compared different sampling approaches using statistical stability measures and evaluated the success rates to construct a surrogate model with relative errors of $<0.1$\%, $<1$\%, and $<10$\%, respectively. The sampling schemes are thoroughly investigated by evaluating the y of surrogate models constructed for various distinct test cases, which represent different problem classes covering low, medium and high dimensional problems. Finally, the sampling schemes are tested on an application example to estimate the sensitivity of the self-impedance of a probe that is used to measure the impedance of biological tissues at different frequencies. We observed strong differences in the convergence properties of the methods between the analyzed test functions.

math.NA↗