DaRUS (University of Stuttgart)
Not a member yet
2037 research outputs found
Sort by
Percentile Intervals in Bayesian Inference are Overconfident
No full textThis dataset demonstrates the difference in calculating percentile Intervals as approximation for Highest Density Intervals (HDI) vs. Highest Posterior Density (HPD). This is demonstrated with extended partial liver resection data (ZeLeR-study, ethical vote: 2018-1246-Material).
The data includes Computed Tomography (CT) liver volume measurements of patients before (POD 0) and after partial hepatectomy. Liver volume was normalized per patient to the preoperative liver volume. was used to screen the liver regeneration courses. The Fujifilm Synapse3D software was used to calculate volume estimates from CT images. The data is structured in a tabular separated value file of the PEtab format
Replication Data for: On the Accurate Estimation of Information-Theoretic Quantities from Multi-Dimensional Sample Data
No full textNon-Parametric Estimation in Information Theory
1. Introduction
This is a repository for our paper on: "On the Accurate Estimation of Information-Theoretic Quantities from Multi-Dimensional Sample Data".
The projects is organizes as follows:
├── analysis_results\
│ ├── plots\
├── data_evaluation\
│ ├── data\
│ ├── notebooks\
│ ├── results\
│ ├── utils\
│ ├── (...) scripts
├── data_generation\
├── README.md
└── .gitignore
2. Installation
Code was written in Python 3.11.5 but should be compatible with later and earlier versions of Python down to Python 3.6. Check the requirements.txt file for any dependency issues.
Usage is recommended by cloning the repository to a local directory and setting up the required environment using venv and pip:
python -m venv .venv
source .venv/Scripts/activate
pip install -r requirements.txt
3. Generating Data
Initially data is generated and stored in the data_evaluation/data directory using the script in the data_generation/ directory. The data for the experiments is stored as an HDF5 database.
From the root directory:
python data_generation/data_generation.py
Note: as the data.hdf5 file is ~123 GB, it is recommended to be locally generated. This process takes about ~12 hrs in an Intel Xeon E5-26280 v2 but shouldn't vary too much in any modern CPU.
4. Conducting an Evaluation
The scripts in the directory data_evaluation/ are used to read the data and perform the experiments. Results are stored in the results/ directory.
Again, from the root directory:
python data_evaluation/eval_bin_entropy.py
All of the names of the scripts have the format eval_{estimator}_{quantity}.py. In total, 12 scripts must be run, tree for each estimator: binning, KDE, numerical integration of KDE and k-NN.
The notebooks/ directory serves as an archive of the development of the workflow to test each estimator. The contents of each notebook are generally the same as the code in the scripts. Log files describe the history of the project.
5. Visualizing Results
The analysis_results directory contains a notebook to create the plots used in the paper, as well as a script to read the log files and calculate the time per iteration of the different experiments.
The plots are generated using the results from the data_evaluation/results directory. Results are read from .hdf5 files.
All results produced using the UNITE Toolbox.</h3
GALÆXI Scaling
No full text
This Dataset contains the results of the scaling test cases of the GALÆXI Paper (Section 4).
The scaling behavior is evaluated using simulations with resolutions ranging from 16384 up to 8.6x109 degrees of freedom using up to 1024 GPUs on JUWELS booster. The provided results contain the raw data of the performance index PID for different loads investigated. The results of the different PIDs are further post-processed to present the strong scaling (speedup over number of devices) and weak scaling (parallel efficiency over number of devices). In case of the strong scaling, results are given with both enabled and disabled streams on the GPU.</p
Code for training and using the droplet segmentation models
No full textThis dataset contains the necessary code for using our spray segmentation model used in the paper, Machine learning based spray process quantification. More information can be found in the README.md
Replication Data for: Automatic Preload Adaptation for Rack-and-Pinion Drives to Maximize Performance and Energy Efficiency
No full textThis dataset contains all experimental data that is shown and referenced within the paper "Automatic Preload Adaptation for Rack-and-Pinion Drives to Maximize Performance and Energy Efficiency".
Electrically preloaded rack-and-pinion drives are typically utilized in machine tools to precisely move heavy loads over long travel distances. In the state of the art, the preload torque is commonly set to a constant value during commissioning. This poses a conflict of objectives between maximizing performance and minimizing energy consumption. To resolve this, this publication presents a novel approach to automatically adapt the preload torque during operation. For this purpose, the established preload control is extended by a simple control law that adapts the preload torque to the current operating state within the permissible limits. This way, higher preload torques are only applied, if the resulting higher system stiffness is beneficial for the drive performance. Otherwise, the preload torque is reduced to save energy.
The automatic preload adaptation is experimentally validated on a test system with industry standard components. This involves both system-theoretical analyses and practical test scenarios.
The data are structured to correspond to the figures in the publication:
Fig. 3 Measurement data: Motor and table velocities for the mechanical frequency response analysis of different preload levels.
Fig. 3 Mechanical frequency response: Mechanical frequency response of the system for different preload levels.
Fig. 5 and 6 Measurement data: Preload torque, motor and table position for the open loop position control frequency response analysis of the three examined configurations.
Fig. 5 Position control frequency response: Open loop position control frequency response of the three examined configurations.
Fig. 6 Preload frequency response: Preload torque of the three examined configurations for the open loop position control frequency response.
Fig. 7 Compliance frequency response: Compliance frequency response of the three examined configurations.
Fig. 7 Measurement data: Linear direct drive force, target and table position for the compliance frequency response analysis of the three examined configurations.
Fig. 7 Preload frequency response: Preload torque of the three examined configurations for the compliance frequency response.
Fig. 8 P2P Motion: Planned trajectory, measured table position and preload torque for the three examined configurations for a point-to-point motion with 2 m/s² acceleration.
Fig. 9 Acceleration comparison: Planned trajectory, measured table position and preload torque for the four examined accelerations with preload adaptation.
Fig. 10 Milling process: Process force applied by the direct drive, path
error and preload torque of the three examined configurations for a simulated milling process.
In addition to the data shown in the publication, this dataset contains supplementary data to Fig. 8 and Fig. 9. In the publication Fig. 8 shows the comparison of the path error and preload torque of the minimum, maximum and adaptive preload exemplarily for a motion with an acceleration of 2 m/s².
The folder Additional PTP Accelerations contains comparable measurement data and the corresponding plots for the other referenced accelerations of 1 m/s², 3 m/s² and 4 m/s². These provide the basis for the data presented in the publication in Table 3
Code and Data for: Better by default: Strong pre-tuned MLPs and boosted trees on tabular data
No full textThis dataset contains code and data for our paper "Better by default: Strong pre-tuned MLPs and boosted trees on tabular data". The main code is provided in pytabkit_code.zip and contains further documentation in README.md and the docs folder. The main code is also provided on GitHub. Here, we additionally provide the data that is generated by the code as well as the plots. See the documentation in docs/source/bench/download_results.md in the main code for instructions on how/when to download which data. The code for the Grinsztajn et al. (2022) benchmark is provided in grinsztajn_benchmarking_code.zip and on GitHub
Data for: A systematic DNS approach to isolate wall-curvature effects in spatially developing boundary layers
No full textSpanwise [z] and temporal [t] averaged turbulent fields of four direct numerical simulations (DNS) of a turbulent boundary layer (TBL) calculated using the code NS3D. Datasets ZPG-C & ZPG-F both have approximately zero pressure gradient along the wall boundary, whereas PG-C & PG-F have a nearly identical non-zero pressure gradient. Both ZPG-C and PG-C have the same curved (C) wall shape and ZPG-F and PG-F are flat (F) plate TBLs. See publication for a full description of the study design. The data are stored in HDF5 containers in which 2D and 1D datasets are present, along with grid coordinates and characteristic freestream conditions of the cases
Setup for High-Speed Synchrotron X-Ray Imaging of Laser Drilling at the DESY
No full textThe image shows the High-Speed Synchrotron X-Ray Imaging setup at the Deutsches Elektronen Synchrotron (DESY) in Hamburg.
It is used to capture X-ray imaging sequences of a laser drilling process with an ultrafast Yb:YAG laser with an output wavelength of 1030 nm. The polarization could be adjusted by two λ/4 and λ/2 waveplates which were placed between the laser source and the focusing f-theta lens.
A sample of thickness t is transmitted by the X-ray beam of diameter about 2 mm in y-direction. At a distance of about 3.5 m behind the sample, a scintillator converts the X-ray beam into visible light, which is captured by a high-speed camera at a frame rate of 1000 fps and a spatial resolution of 856 pixels/mm.
In addition, the measurement beam of an optical coherence tomography (OCT) is aligned coaxially with the processing laser beam using a dichroitic mirror.</p
DuMux 3.9.0
No full textRelease 3.9.0 of DuMux, DUNE for Multi-{Phase, Component, Scale, Physics, ...} flow and transport in porous media. DuMux is a free and open-source simulator for flow and transport processes in and around porous media. It is based on the Distributed and Unified Numerics Environment DUNE
Supplementary Material for Uncertainty-aware Spectral Visualization
No full textIn this supplemental material, we provide supplemental information (PDF document with derivations of the results presented in the paper and two additional use cases) and the supplementary video for uncertainty-aware spectral analysis. We model an uncertain time series as a multivariate Gaussian process. We propagate the uncertainty and explicitly compute the probability distribution in the spectral domain. In the video, we use our interactive visual analysis tool to analyze these distributions. This material complements the paper on Uncertainty-aware Spectral Visualization