Find research datasets worth reusing

Suplementary Material for Direct hydrodynamic measurements at the upper shoreface of a sandy beach in Paraná - Brazil

<p>ADCP data accompanying the manuscript entitled "Direct hydrodynamic measurements at the upper shoreface of a sandy beach in Paraná - Brazil" to be published in Ocean and Coastal Research<strong> (</strong>ISSN (<em>online</em>): 2675-2824)</p>

High Tide or Rip-Tide? GJ 486b JWST Transits with NIRSpec G395H - Supplementary Material

<p>Supplementary material for '<em>High Tide or Riptide on the Cosmic Shoreline? A Water-rich Atmosphere or Stellar Contamination for the Warm Super-Earth GJ 486b from JWST Observations</em>' by Moran & Stevenson et al., ApJL (2023).</p> <p>This repository contains four categories of data products:</p> <ol> <li><strong>White light curves</strong> — White light curves produced for Visit 1 and 2, for detectors NRS1 and NRS2 via the <a href="https://ui.adsabs.harvard.edu/abs/2023Natur.614..659R/abstract">FIREFLy pipeline</a> and the <a href="https://github.com/kevin218/Eureka">open-source codes Eureka</a>! and <a href="https://github.com/JamesKirk11/Tiberius">Tiberius</a></li> <li><strong>Transmission spectra</strong> — Reduced transmission spectra from three data reduction codes (Eureka!, FIREFLy, and Tiberius) for both G395H visits.</li> <li><strong>Stellar spectra and models </strong>— Flux calibrated observed stellar spectra of GJ 486 from each G395H visit, along with best-fit PHOENIX stellar models.</li> <li><strong>Atmospheric models</strong> — Atmospheric forward models used to interpret the data, produced from the open source <a href="https://natashabatalha.github.io/picaso/">PICASO</a> and <a href="https://github.com/mrline/CHIMERA">CHIMERA</a> codes.</li> </ol> <p>For any additional data requests or questions, please contact: sarahemoran@arizona.edu or kevin.stevenson@jhuapl.edu</p> <p> </p>

opencc-by-4.0Jul 2023View details →

The response of silicon carbide composites to He ion implantation and ramifications for use as a fusion reactor structural material

<p>Raw datasets for the publication "The response of silicon carbide composites to He ion implantation and ramifications for use as a fusion reactor structural material".</p>

opencc-by-nc-sa-4.0Dec 2023View details →

A dataset of 1600 images extracted from 5 cm RGB orthophotos for the classification of 12 classes of roofing materials

<p>This dataset contains a collection of 1601 image tiles of 64x64 pixels (3.2x3.2m²) annotated for 12 roofing materials. These tiles were extracted from 5 cm RGB orthophotos acquired by the city of Namur (Belgium) in 2017. The additional data used to create this dataset are (a) a Namur roof section mask, and (b) a set of 1601 material samples acquired using stratified random sampling. The tiles were obtained as follows: the centroid of each roof section containing a sample is used to extract tiles. A size of 64x64 pixels has been chosen so that a tile contains information for only one roof section, in order to learn only the colour and texture of the roof materials. This also avoids adding information outside the given roof section. The tiles are thus extracted for each orthophoto spectral band and labelled with the identifier of the class of roofing materials to which they belong. Here are the 12 material classes considered, preceded by their labels:</p> <p>0- Solar panels<br>1- Brown tiles<br>2- Orange tiles<br>3- Black tiles<br>4- Dark membranes<br>5- White membranes<br>6- Slates containing asbestos<br>7- Slates without asbestos<br>8- Corrugated asbestos-cement sheets<br>9- Gravel<br>10- Vegetation<br>12- Metals</p> <p>There are approximately 140 tiles by material class except for the vegetated roof sections (class 10) which contains only 47 samples due to its rarety.</p> <p>The dataset contains 1 folder for each spectral band. Each folder contains 1601 thumbnails in tif format named as follows:</p> <p><strong>img[tile id]_[class label].tif</strong></p> <p>It is suggested to apply pre-processing to these images as done by<a href="https://doi.org/10.1109/jurse57346.2023.10144142"> Wyard et al. (2023).</a></p>

Supplementary Material: The Importance of Optical Wavelength Data on Atmospheric Retrievals of Exoplanet Transmission Spectra

<p>Supplementary material for "The Importance of Optical Wavelength Data on Atmospheric Retrievals of Exoplanet Transmission Spectra" DOI: <a href="https://ui.adsabs.harvard.edu/link_gateway/2024arXiv240307801F/doi:10.48550/arXiv.2403.07801" target="_blank" rel="noreferrer noopener">10.48550/arXiv.2403.07801</a></p> <p>Contents of this record:</p> <ul> <li>The retrieved atmospheric parameters for the population (see supplementary_material.pdf).</li> <li>The retrieval statistics per planet for the wavelength ranges 0.3-4.5, 0.6-4.5, and 1.1-4.5 microns (see supplementary_material.pdf).</li> <li>Planet specific retrieved spectra for the wavelength ranges 0.3-4.5, 0.6-4.5, and 1.1-4.5 microns.</li> <li>Retrieved parameter cornerplots per planet for the wavelength ranges 0.3-4.5, 0.6-4.5, and 1.1-4.5 microns.</li> </ul> <p>(NOTE:  the retrieval model and priors for the results displayed in this record are specified in tables 2 and 3 of the paper.)</p>

opencc-by-4.0Oct 2023View details →

START video for the Raw Materials Week 2023

<p>START video for the Raw Materials Week 2023 (Brussels, 13-17 November, 2023), including an overview on the project and Starty on the Critical Raw Materials issues in the EU and the START contribution on that!</p>

opencc-zeroDec 2023View details →

Supplementary material from: Deep divergences among inconspicuous clades of Epipedobates poison frogs

<p>Poison frogs (<em>Dendrobatidae</em>) are famous for their aposematic species, having a combination of diverse color patterns and defensive skin toxins, yet most species in this family are inconspicuously colored and considered non-aposematic. <em>Epipedobates</em> is among the youngest genus-level clades of <em>Dendrobatidae </em>that includes both aposematic and inconspicuous species. Using Sanger-sequenced mitochondrial and nuclear markers, we demonstrate deep genetic divergences among inconspicuous species of <em>Epipedobates</em> but relatively shallow genetic divergences among conspicuous species. Our phylogenetic analysis includes broad geographic sampling of the inconspicuous lineages typically identified as <em>E. boulengeri </em>and <em>E. espinosai</em>, which reveals two putative new species, one in west-central Colombia (<em>E.</em> sp. 1) and the other in north-central Ecuador (<em>E. </em>aff. <em>espinosai</em>). We conclude that <em>E. darwinwallacei</em> is a junior subjective synonym of <em>E. espinosai</em>. We also clarify the geographic distributions of inconspicuous <em>Epipedobates </em>species including the widespread <em>E. boulengeri.</em> We provide a qualitative assessment of the phenotypic diversity in each nominal species, with a focus on the color and pattern of inconspicuous species. We conclude that <em>Epipedobates</em> contains eight known valid species, six of which are inconspicuous. A relaxed molecular clock analysis suggests that the most recent common ancestor of <em>Epipedobates</em> is ~11.1 million years old, which nearly doubles previous estimates. Last, genetic information points to a center of species diversity in the Chocó at the southwestern border of Colombia with Ecuador.</p>

Supplementary Materials - Integrating Spatial Analyses and Microbotanical Remains: A Methodological Approach for Investigating Plant Processing Activities and Domestic Spaces at Neolithic Çatalhöyük.

<p><strong>Supplementary Materials I, II, and III - Integrating Spatial Analyses and Microbotanical Remains: A Methodological Approach for Investigating Plant Processing Activities and Domestic Spaces at Neolithic Çatalhöyük.</strong></p> <p> </p> <p><strong>Supplementary I</strong> contains the raw data for the microbotanical analyses on Buildings 80 and 131. It consists of laboratory track sheets and the raw counts of phytoliths and starch grains from each building.</p> <p><strong>Supplementary II</strong> contains: a) A detailed description of all the starch grain typologies encountered in this research along with references from modern reference collections; b) A visual example of each of the phytolith morphotypes identified and c) The variogram and the Kriging error variances.</p> <p><strong>Supplementary III</strong> refers to the R code, building shape files (Mask), and spatial data utilized to produce the Kriging and IDW stipital objects.</p>

opencc-byJun 2021View details →

HighPoly three heads ganesha With Materials

Hey! since a few days i comeback to substance designer to try to create some bunch of diferents basic materials for my projects, now i saw a new workflow with zbrush and i will try it the next weeks Source: Objaverse 1.0 / Sketchfab

openmit-licenseDec 2023View details →

Supplementary material for the publication: Deep Learning of Crystalline Defects from TEM images: A Solution for the Problem of "Never Enough Training Data"

Open the record for dataset details and reuse information.

Supplementary material data for: Unstable environmental conditions constrain the fine-tune between opsin sensitivity and underwater light in an Amazon forest stream fish

<p>Visual adaptations can stem from variations in amino acid composition, chromophore utilization, and differential opsin gene expression levels, enabling individuals to adjust their light sensitivity to environmental lighting conditions. In stable environments, adaptations often involve amino acid substitutions, whereas in unstable conditions, differential gene expression may be a more relevant mechanism. Amazon forest streams present diverse underwater lighting conditions and experience short-term water colour fluctuations. In these environments, it is less likely for genetic and amino acid sequences to undergo modifications that tailor opsin proteins to the prevailing lighting conditions, particularly in species having several copies of the same gene. The sailfin tetra, <em>Crenuchus spilurus</em>, inhabits black and clear water Amazon forest streams. The long wavelength sensitivity (LWS) is an important component for foraging and courtship. Here, we investigated LWS opsin genes in the <em>sailfin tetra</em>. Three copies of LWS1 and two copies of LWS2 genes were found. The maximum absorbance wavelength (λmax) estimated from the amino acid sequences of LWS1 genes exhibited variation among the different copies. In contrast, the copies of LWS2 genes showed identical expected λmax values. Although the amino acid positions affecting λmax varied among LWS genes, they remained consistent among populations living in different water colours. The relative expression levels of LWS genes differed between gene copies. While not formally tested, our results suggest that in fluctuating environments, visual adaptations may primarily stem from alterations in gene expression profiles and/or chromophore usage rather than precise genetic tuning of protein light sensitivity to environmental lighting conditions.</p>

Quantum Material-Based Self-Propelled Microrobots for the Optical "On-the-Fly" Monitoring of DNA

<p>Quantum dot-based materials have been found to be excellent platforms for biosensing and bioimaging applications. Herein, self-propelled microrobots made of graphene quantum dots (GQD–MRs) have been synthesized and explored as unconventional dynamic biocarriers toward the optical “on-the-fly” monitoring of DNA. As a first demonstration of applicability, GQD–MRs have been first biofunctionalized with a DNA biomarker (i.e., fluorescein amidite-labeled, FAM-L) via hydrophobic π-stacking interactions and subsequently exposed toward different concentrations of a DNA target. The biomarker–target hybridization process leads to a biomarker release from the GQD–MR surface, resulting in a linear alteration in the fluorescence intensity of the dynamic biocarrier at the nM range (1–100 nM, <em>R</em><sup>2</sup> = 0.99), also demonstrating excellent selectivity and sensitivity, with a detection limit as low as 0.05 nM. Consequently, the developed dynamic biocarriers, which combine the appealing features of GQDs (e.g., water solubility, fluorescent activity, and supramolecular π-stacking interactions) with the autonomous mobility of MRs, present themselves as potential autonomous micromachines to be exploited as highly efficient and sensitive “on-the-fly” biosensing systems. This method is general and can be simply customized by tailoring the biomarker anchored to the GQD–MR’s surface.</p>

Supplementary material from: Alpine extremophytes in evolutionary turmoil: Complex diversification patterns and demographic responses of a Halophilic grass in a Central Asian biodiversity hotspot

<p>Diversification and demographic responses are key processes shaping species evolutionary history. Yet we still lack a full understanding of ecological mechanisms that shape genetic diversity at different spatial scales upon rapid environmental changes. In this study, we examined genetic differentiation in an extremophilic grass <em>Puccinellia pamirica</em> and factors affecting its population dynamics among the occupied hypersaline alpine wetlands on the arid Pamir Plateau in Central Asia. Using genomic data, we found evidence of fine-scale population structure and gene flow among the localities established across the high-elevation plateau as well as fingerprints of historical demographic expansion. We showed that an increase in the effective population size could coincide with the Last Glacial Period, which was followed by the species demographic decline during the Holocene. Geographic distance plays a vital role in shaping the spatial genetic structure of <em>P. pamirica</em> alongside with isolation-by-environment and habitat fragmentation. Our results highlight a complex history of divergence and gene flow in this species-poor alpine region during the Late Quaternary. We demonstrate that regional climate specificity and a shortage of nonclimate data largely impede predictions of future range changes of the alpine extremophile using ecological niche modeling. This study emphasizes the importance of fine-scale environmental heterogeneity for population dynamics and species distribution shifts.</p>

opencc-by-4.0Nov 2023View details →

Heterolayered carbon allotrope architectonics via multi-material 3D printing for advanced electrochemical devices

<p>3D printing has become a powerful technique in electrochemistry for fabricating electrodes, thanks to readily available conductive nanocomposite filaments, such as those based on carbon fillers (i.e., carbon nanotubes (CNTs) or carbon black (CB)) within an insulating polymeric matrix like polylactic acid (PLA). Inspired by inorganic heterostructures that enhance the functional characteristics of nanomaterials, we fabricated hetero-layered 3D printed devices based on carbon allotropes using a layer-by-layer assembly approach. The heterolayers were customised through the alternate integration of different carbon allotrope filaments via a multi-material 3D printing technique, allowing for a time-effective method to enhance electrochemical performance. As a first demonstration of applicability, CNT/PLA and CB/PLA filaments were utilised to construct ordered hetero-layered carbon-based electrodes. This contrasts with conventional methods where various carbon species are mixed in the same composite-based filament used for building electrochemical devices. Multi-material 3D-printed carbon electrodes exhibit improved electrochemical performance in energy conversion (e.g., hydrogen evolution reaction or HER) and sensing applications (e.g., ascorbic acid detection) compared to single-material electrodes. This work paves the way for manufacturing advanced 3D-printed heterolayered electrodes with enhanced electrochemical activity through multi-material 3D printing technology.</p>

Supplementary material: Machine learning and phylogenetic models identify predictors of genetic variation in Neotropical amphibians

<p><strong>Aim: </strong>Intraspecific genetic variation is key for adaptation and survival in changing environments and is known to be influenced by many factors, including population size, dispersal, and life history traits. We investigated genetic variation within Neotropical amphibian species to provide insights into how natural history traits, phylogenetic relatedness, climatic, and geographic characteristics can explain intraspecific genetic diversity.</p> <p><strong>Location:</strong> Neotropics.</p> <p><strong>Taxon:</strong> Amphibians.</p> <p><strong>Methods: </strong>We assembled datasets using open-access databases for natural history traits, genetic sequences, phylogenetic trees, climatic, and geographic data. For each species, we calculated overall nucleotide diversity (<em>π</em>) and tested for isolation by distance (IBD) and isolation by environment (IBE). We then identified predictors of <em>π</em>, IBD, and IBE using Random Forest (RF) regression or RF classification. We also fitted phylogenetic generalized linear mixed models (PGLMMs) to predict<em> π</em>, IBD, and IBE.</p> <p><strong>Results:</strong> We compiled 4,052 mitochondrial DNA sequences from 256 amphibian species (230 frogs and 26 salamanders), georeferencing 2,477 sequences from 176 species that were not linked to occurrence data. RF regressions and PGLMMs were congruent in identifying range size and precipitation (σ) as the most important predictors of <em>π</em>, influencing it positively. RF classification and PGLMMs identified minimum elevation as an important predictor of IBD; most species without IBD tended to occur at higher elevations. Maximum latitude and precipitation (σ) were the best predictors of IBE, and most species without IBE occur at lower latitudes and in areas with more variable precipitation.</p> <p><strong>Main conclusions: </strong>This study identified predictors of genetic variation in Neotropical amphibians using both machine learning and phylogenetic methods. This approach was valuable to determine which predictors were congruent between methods. We found that species with small ranges or living in zones with less variable precipitation tended to have low genetic diversity. We also showed that Western Mesoamerica, Andes, and Atlantic Forest biogeographic units harbor high diversity across many species that should be prioritized for protection. These results could play a key role in the development of conservation strategies for Neotropical amphibians.</p>

Moth proboscis morphology material properties raw data

<p>Insects have evolved unique structures that host a diversity of material and mechanical properties, and the mouthparts (proboscis) of butterflies and moths (Lepidoptera) are no exception. Here, we examined proboscis morphology and material properties from several previously unstudied moth lineages to determine if they relate to flower-visiting and non-flower-visiting feeding habits. Scanning electron microscopy and 3D imaging were used to study proboscis morphology and assess surface roughness patterns on the galeal surface, respectively. Confocal laser scanning microscopy was used to study patterns of cuticular autofluorescence, which was quantified with color analysis software. We found that moth proboscises display similar color and morphological patterns in relation to these feeding habits as those previously described for flower and non-flower-visiting butterflies. The distal region of proboscises of non-flower visitors is brush-like for augmented capillarity and exhibited blue autofluorescence, indicating the possible presence of resilin and increased flexibility. Flower visitors have smoother proboscises and show red autofluorescence, an indicator of high sclerotization, which is adaptive for floral tube entry. We propose the lepidopteran proboscis as a model structure for understanding how insects have evolved a suite of morphological and material adaptations to overcome the challenges of acquiring fluids from diverse sources.</p>

Negative Complement of a Set of Vulnerability-Fixing Commits: Supplementary Material

<div> <div><span>EASE 2024 - Industry track</span></div> <br> <div><span>This archive contains accompanying materials for the paper: </span></div> <br> <div><span>Rocío Cabrera Lozoya, Antonino Sabetta, Tommaso Aiello. "Negative Complement of a Set of Vulnerability-Fixing Commits: Method and Dataset" </span></div> <br> <div><span>submitted to the Industry track at EASE 2024 - https://conf.researchr.org/track/ease-2024/ease-2024-industry</span></div> <br> <div><span>It contains the following folders and files:</span></div> <br> <div><span>-</span><span> data</span></div> <div><span> </span><span>-</span><span> commit_pairs_final.csv : Dataset of 534 commit pairs corresponding to a positiive (security-relevant) commit and a negative sample obtained by the approach described in the paper.</span></div> <div><span> </span><span>-</span><span> single_positive.csv : Contains a single positive instance (taken from the MSR2019 dataset) which can be used to test the generate_negative_complement.py script.</span></div> <div><span>-</span><span> scripts</span></div> <div><span> </span><span>-</span><span> generate_negative_complement.py : Takes in a single .java file and obfuscates its developer-defined identifiers.</span></div> <div><span> </span><span>-</span><span> obfuscate_java_file.py : Generates a negative complement for a security-relevant dataset. The ouput is written to couple_dataset.csv.</span></div> <div><span> </span><span>-</span><span> requirements.txt : Requirements needed in the virtual environment to successfully run the previous scripts.</span></div> </div>

opencc-by-4.0Jan 2024View details →

Data of "Self-consistency Reinforced minimal Gated Recurrent Unit for surrogate modeling of history-dependent non-linear problems: application to history-dependent homogenized response of heterogeneous materials"

<h1>Development of the Self-Consistency reinforced Minimum Recurrent Unit (SC-MRU)</h1> <p>This directory contains the data and algorithms generated in publication<sup><a href="#fn-1-5292">1</a></sup></p> <h2>Table of Contents</h2> <ol> <li><a href="#dependencies-and-prerequisites">Dependencies and Prerequisites</a></li> <li><a href="#structure-of-repository">Structure of Repository</a></li> <li><a href="#part-1-data-preparation">Part 1: Data preparation</a></li> <li><a href="#part-2-rnn-training">Part 2: RNN training</a></li> <li><a href="#part-3-multiscale-analysis">Part 3: Multiscale analysis</a></li> <li><a href="#part-4-reproduce-paper1-figures">Part 4: Reproduce paper[^1] figures</a></li> </ol> <h2>Dependencies and Prerequisites</h2> <p> </p> <ul> <li> <p>Python, pandas, matplotlib, texttabble and latextable are pre requisites for visualizing and navigating the data.</p> </li> <li> <p>For generating mesh and for vizualization, gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) is required.</p> </li> <li> <p>For running simulations, cm3Libraries (<a href="http://www.ltas-cm3.ulg.ac.be/openSource.htm" target="_blank" rel="nofollow noreferrer noopener">http://www.ltas-cm3.ulg.ac.be/openSource.htm</a>) is required.</p> </li> </ul> <h3>Instructions using apt & pip3 package manager</h3> <p>Instructions for Debian/Ubuntu based workstations are as follows.</p> <h3>python, pandas and dependencies</h3> <div> <pre><code> sudo apt install python3 python3-scipy libpython3-dev python3-numpy python3-pandas</code></pre> </div> <h3>matplotlib, texttabble and latextable</h3> <div> <pre><code> pip3 install matplotlib texttable latextable</code></pre> </div> <h3>Pytorch (only for run with cm3Libraries)</h3> <ul> <li>Without GPU</li> </ul> <div> <pre><code> pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu</code></pre> </div> <ul> <li>With GPU</li> </ul> <div> <pre><code> pip3 install torch torchvision torchaudio</code></pre> </div> <h3>Libtorch (for compiling the cells)</h3> <ul> <li>Without GPU: In a local directory (e.g. <code>~/local</code> with <code>export TORCHDIR=$HOME/local/libtorch</code>)</li> </ul> <div> <pre><code> wget https://download.pytorch.org/libtorch/cpu/libtorch-shared-with-deps-2.1.1%2Bcpu.zip unzip libtorch-shared-with-deps-2.1.1%2Bcpu.zip</code></pre> </div> <ul> <li>With GPU: In a local directory (e.g. <code>~/local</code> with <code>export TORCHDIR=$HOME/local/libtorch</code>)</li> </ul> <div> <pre><code> wget https://download.pytorch.org/libtorch/cu121/libtorch-shared-with-deps-2.1.1%2Bcu121.zip unzip libtorch-shared-with-deps-2.1.1+cu121.zip</code></pre> </div> <h2>Structure of Repository</h2> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/All_Path_Res">All_Path_Res</a>: results of the direct numerical simulations used as training and testing data, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/ConstRVE">ConstRVE</a>: script to run direct numerical finite element simulations, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale</a>: scripts to run and visualise the multiscale analyses, see details in <a href="#part-3-multiscale-analysis">Part 3: Multiscale analysis</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU</a>: implementation of the RNN and scripts to train them, see details in <a href="#part-2-rnn-training">Part 2: RNN training</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a>: scripts to collect, normalise and truncate the RVEs direct simulation results as training and testing data, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>. The director also contained the storred processed data used in <sup><a href="#fn-1">1</a></sup>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a>: scripts to generate the different loading paths for the direct numerical simulations used as training and testing data, see details in <a href="#part-1-data-preparation">Part 1: Data preparation</a>.</li> </ul> <h2>Part 1: Data preparation</h2> <h3>Generate the loading paths</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/testGenerationData.py">TrainingPaths/testGenerationData.py</a> is used to generate random walk paths, with the options <ul> <li><code>Rmax = 0.11</code> # bound on the final Green Lagrange strain</li> <li><code>TimeStep = 1.</code> # in second</li> <li><code>EvalStep = [1e-4,5e-3]</code> #Bounds on the Green Lagrange increments</li> <li><code>Nmax = 2500</code> #maximum length of the sequence</li> <li><code>k = 4000</code> # number of path to generate</li> <li>The path are storred by default in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a>. The path has to be existing before launching the script. You can change the name in line 123 <code>saveDir = '../ConstRVE'+'/Paths/'</code>.</li> <li>Examples of generated paths can be found in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/ConstRVE/PathsExamples">ConstRVE/PathsExamples/</a></li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a> is</li> </ul> </li> </ul> <div> <pre><code>(mkdir ../ConstRVE/Paths) #if needed python3 testGenerationData.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/generationData_Cyclic.py">TrainingPaths/generationData_Cyclic.py</a> is used to generate random cylic paths, with the options <ul> <li><code>Rmax = [np.random.uniform(0.,0.04),np.random.uniform(0.,0.06),np.random.uniform(0.0,0.09),0.12]</code> # bound on the final Green Lagrange strain is random</li> <li><code>TimeStep = 1.</code> # in second</li> <li><code>EvalStep = [1e-4,5e-3]</code> #Bounds on the Green Lagrange increments</li> <li><code>Nmax = 2500</code> #maximum length of the sequence</li> <li><code>k = 2000</code> # number of path to generate</li> <li>The path are stored by default in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a>. You can change the name in line 123 <code>saveDir = '../ConstRVE'+'/Paths/'</code>.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a> is</li> </ul> </li> </ul> <div> <pre><code>(mkdir ../ConstRVE/Paths) #if needed python3 generationData_Cyclic.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/countPathLength.py">TrainingPaths/countPathLength.py</a> gives average, minimum and maximum lengths of the generated paths and the distribution of the <code>\Delta R</code>. By default the paths are read in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a> but the directory can be given as an argument. The file can be used to read <ul> <li>either the generated loading paths</li> </ul> </li> </ul> <div> <pre><code> python3 countPathLength.py '../ConstRVE/PathsExamples'</code></pre> </div> <ul> <li> <ul> <li>or the results of the <a href="#generate-the-rves-direct-simulation-results">simulations</a></li> </ul> </li> </ul> <div> <pre><code> python3 countPathLength.py '../All_Path_Res/Path_Res9'</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingPaths/graphData.py">TrainingPaths/graphData.py</a> generates illustrations from randomly picked paths in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a> and generate png figures.</li> </ul> <h3>Generate the RVEs direct simulation results</h3> <ul> <li>Uses the loading paths existing in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Paths">ConstRVE/Paths/</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/rve.geo">ConstRVE/rve.geo</a> is the RVE geometry file that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/rve.msh">ConstRVE/rve.msh</a> is the RVE mesh file that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/utilsFunc.py">ConstRVE/utilsFunc.py</a> contains python tools to be used.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/ConstRVE/Rve_withoutInternalVars.py">ConstRVE/Rve_withoutInternalVars.py</a> is used to run all the RVE simulations: <ul> <li>This requires cm3Libraries (<a href="http://www.ltas-cm3.ulg.ac.be/openSource.htm" target="_blank" rel="nofollow noreferrer noopener">http://www.ltas-cm3.ulg.ac.be/openSource.htm</a>).</li> <li>All the ouptus are stored in <code>All_Path_Res/Path_Res12</code>, you can change the name in line 71 <code>Path_Res = '../All_Path_Res/Path_Res12/'</code>. The results of RVE simulations are saved as the sequence (one configuration per line) of the Green-Lagrange strains and Second Piola-Kirchhoff stress (in column). One example can be found in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_Res1/data_path1000.csv">All_Path_Res/Path_Res1/data_path1000.csv</a>.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/ConstRVE">ConstRVE</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 Rve_withoutInternalVars.py</code></pre> </div> <h3>Collect, normalised and truncate the RVEs direct simulation results as training and testing data</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/CheckNanData.py">TrainingData/CheckNanData.py</a> is used to check the integrity of the direct numerical simulations results <ul> <li>DNS results are read from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res1">All_Path_Res/Path_res1</a> to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res11">All_Path_Res/Path_res11</a> subdirectories.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 CheckNanData.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/CollectData.py">TrainingData/CollectData.py</a> is used to gather all the direct numerical simulations results <ul> <li>DNS results are read from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res1">All_Path_Res/Path_res1</a> to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/All_Path_Res/Path_res11">All_Path_Res/Path_res11</a> subdirectories. It can be changed in line 31 <code>for ll in range(11):</code>.</li> <li>It saves the bounds and raw data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Bounds_GS">TrainingData/Processed_Data/Bounds_GS</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Origin_GS">TrainingData/Processed_Data/Origin_GS</a>, respectively.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 CollectData.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Normalization.py">TrainingData/Normalization.py</a> is used to normalise the gathered direct numerical simulations results <ul> <li>Bounds and raw data are read from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Bounds_GS">TrainingData/Processed_Data/Bounds_GS</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Origin_GS">TrainingData/Processed_Data/Origin_GS</a>, respectively.</li> <li>It saves the normalized training (75%) and testing (25%) data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Train">TrainingData/Processed_Data/Normalized_GS_Train</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Test">TrainingData/Processed_Data/Normalized_GS_Test</a>, respectively.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 Normalization.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/GaussCollectData.py">TrainingData/GaussCollectData.py</a> is an alternative using Gaussian normaliation and is used to gather all the direct numerical simulations results.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/GaussNormalization.py">TrainingData/GaussNormalization.py</a> is an alternative to normalise following a Gaussian the gathered direct numerical simulations results.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Data_Padding.py">TrainingData/Data_Padding.py</a> is used to pad and trim the normalised data <ul> <li>Normalized training (75%) and testing (25%) data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Train">TrainingData/Processed_Data/Normalized_GS_Train</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data/Normalized_GS_Test">TrainingData/Processed_Data/Normalized_GS_Test</a>, respectively.</li> <li>The final length of the sequence (including zero padding and trimming) is given by <code>N = 200</code>.</li> <li>It saves the trimmed and padded normalised training and testing data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data">TrainingData/Processed_Data/</a> as <code>'GS_Train'N</code> and <code>'GS_Test'N</code>, respectively.</li> <li>The command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingData">TrainingData</a> directory is</li> </ul> </li> </ul> <div> <pre><code> python3 Data_Padding.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Tool.py">TrainingData/Tool.py</a> is a list of function used to normalise dat.</li> </ul> <h2>Part 2: RNN training</h2> <h3>Available rnn</h3> <ul> <li>The different cells are in the following directories <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SMRU">SC_MRU/MGRU/NNW_SMRU</a>: Neural network with SMRU recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SCMRU_T">SC_MRU/MGRU/NNW_SCMRU_T</a>: Neural network with SC-MRU-T recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_MGRU">SC_MRU/MGRU/NNW_MGRU</a>: Neural network with orginal MGRU recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_MGRU_M">SC_MRU/MGRU/NNW_MGRU_M</a>: Neural network with modified MGRU recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/FNN_LeakyReLU">SC_MRU/FNN_LeakyReLU/NNW_<code>X</code>Fw</a>: Neural network with SC-MRU-I recurrent cell using <code>X</code> Feed Forward non-linear transition layers.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/Quadratic_NLT">SC_MRU/Quandratic_NLT/NNW_<code>X</code>Q, NNW_Q_Fw, NNW_Fw_Q</a>: Neural network with SC-MRU-I recurrent cell using <code>X</code> quadratic or mixed feed-forward-quadratic non-linear transition layers.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/ReferenceRNN">SC_MRU/ReferenceRNN/NNW_<code>X</code>Layers</a>: Neural network with SC-LMSC recurrent cell using <code>X</code> non-linear transition layers.</li> </ul> </li> </ul> <h3>Compile the neural network models</h3> <ul> <li>In the adequate directory, e.g. <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SMRU">SC_MRU/MGRU/NNW_SMRU</a> for the Neural network with SMRU recurrent cell.</li> </ul> <div> <pre><code> cd build rm -rf * cmake -DCMAKE_PREFIX_PATH=$TORCHDIR .. make</code></pre> </div> <ul> <li>This create the RNN_<code>CELL</code> model in the build directory</li> </ul> <h3>Train the neural network models</h3> <ul> <li>In the adequate directory, e.g. <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/MGRU/NNW_SMRU">SC_MRU/MGRU/NNW_SMRU</a> for the Neural network with SMRU recurrent cell.</li> <li>Requires to have <a href="#compile-the-neural-network-models">compiled</a> the RNN model.</li> <li>The file <code>Train.py</code> <ul> <li>Uses the RNN_<code>CELL</code> model compiled in the <code>build</code> directory.</li> <li>Uses the trimmed and padded normalised training and testing data in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/TrainingData/Processed_Data">TrainingData/Processed_Data/</a> as <code>'GS_Train'N</code> and <code>'GS_Test'N</code>, respectively.</li> <li>Can be modified to use the requested ratios of training and testing sequences of different lengths to prepate the mini-baches, e.g.: <ul> <li><code>ratio = [0.2,0.02,0.6,0.06,0.01,0.9]</code></li> <li><code>PathIn1 = ['../../../TrainingData/Processed_data/GS_Train200','../../../TrainingData/Processed_data/GS_Train2500','../../../TrainingData/Processed_data/GS_Test200','../../../TrainingData/Processed_data/GS_Test2500']</code></li> <li><code>PathIn2 = ['../../../TrainingData/Processed_data/GS_Test200','../../../TrainingData/Processed_data/GS_Test400','../../../TrainingData/Processed_data/GS_Test2500','../../../TrainingData/Processed_data/GS_Test400']</code></li> </ul> </li> <li>Saves the mini-batches in <ul> <li><code>PathOut = "TrainingData.pt"</code></li> </ul> </li> <li>Saves the model <code>module-checkpoint.pt</code> and <code>module-checkpoint-optimizer.pt</code>, and loss evolution <code>Loss.txt</code> in <ul> <li>Module/H<code>n</code> with <code>n</code> the number of hiden variables.</li> <li>Output directory can be changes in NNW_<code>CELL</code>.cpp of the cell name <code>CELL</code> (and recompiling).</li> <li>Warm start can be disabled in by commenting <code>torch::load(net, "./Module/H120/module-checkpoint_CM0.pt");</code> and <code>torch::load(optimizer, "./Module/H120/module-optimizer-checkpoint_CM0.pt");</code> in NNW_<code>CELL</code>.cpp of the cell name <code>CELL</code> (and recompiling).</li> <li>Is executed with</li> </ul> </li> </ul> </li> </ul> <div> <pre><code> python3 Train.py</code></pre> </div> <h3>Convert trained c++ models for pyTorch (in view of multiscale simulations)</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Save_RU.py">SC_MRU/CheckLoss/RU.py</a>: functions used to read and use the rnn models.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Save_RNN_script.py">SC_MRU/CheckLoss/Save_RNN_script.py</a>: <ul> <li>Converts c++ trained models to pyTorch models</li> <li>Reads the trained models in SC_MRU/CELL_<code>Kind</code>/NNW_<code>CELL</code>/Module/H<code>N</code>/module.pt</li> <li>Save the converted models to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> and to <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/Model">MultiScale/Model/</a></li> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells <code>cell=["SC_MRU_T","SC_MRU_I","SMRU"]</code>.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is</li> </ul> </li> </ul> <div> <pre><code>python3 Save_RNN_script.py</code></pre> </div> <h3>Vizualize loss evolution and testing results</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Data_prepare.py">SC_MRU/CheckLoss/Data_prepare.py</a> contains fucntion used for testing.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/LossVS_insertN.py">SC_MRU/CheckLoss/LossVS_insertN.py</a> tests the effect of testing data augmentation: <ul> <li>Requires to have <a href="#compile-the-neural-network-models">compiled</a> the RNN models.</li> <li>Cell type and augmentation kind can be modified: <ul> <li><code>repeat = 10</code> #number of tests</li> <li><code>dataType ='random'</code> #'even' or 'random'</li> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells ```cell=["SC_MRU_T","SC_MRU_I","SMRU"]``</li> <li><code>reevaluate= False</code> # <code>False</code> to use the saved loss values and <code>True</code> to evaluate the loss values</li> <li><code>fastshifting=False</code> # <code>False</code> to vizualize before fast shifting (Fig. 11) and <code>True</code> after (Fig. 12). When <code>reevaluate== True</code>, this has no effect: the training with fast-shifting or not has to be done manually, see <a href="#train-the-neural-network-models">details</a>.</li> </ul> </li> <li>Generates new Loss_<code>CELL</code>.txt and TrainingData.pt files in case <code>reevaluate== True</code>, read the existing one if <code>reevaluate== False</code>.</li> <li>Generates Figs. 11 and 12 from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a></li> </ul> </li> </ul> <div> <pre><code> python3 LossVS_insertN.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Plot_GS.py">SC_MRU/CheckLoss/Plot_GS.py</a> test the different NNW on RVE testing paths: <ul> <li>Requires to have <a href="#compile-the-neural-network-models">compiled</a> the RNN models.</li> <li>Cell type and augmentation kind can be modified: <ul> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells ```cell=["SC_MRU_T","SC_MRU_I","SMRU"]``</li> </ul> </li> <li>Generate Figs. 13, 14 and 15 from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a></li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 Plot_GS.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/CheckLoss/Plot_GS_step.py">SC_MRU/CheckLoss/Plot_GS_step.py</a> tests the different pyTorch NNWs on the RVE testing paths: <ul> <li>Requires to have <a href="#convert-trained-c-models-for-pytorch-in-view-of-multiscale-simulations">converted</a> the RNN models.</li> <li>Cell type and augmentation kind can be modified: <ul> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells ```cell=["SC_MRU_T","SC_MRU_I","SMRU"]``</li> </ul> </li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 Plot_GS_step.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_HiddenV.py">SC_MRU/PlotLoss_HiddenV.py</a> shows the loss evolution for the different number of hidden variables --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- for the SMRU cell: <ul> <li>Generates Fig. 8.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_HiddenV.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_3MRU.py">SC_MRU/PlotLoss_3MRU.py</a> shows the loss evolution for the different recurrent units --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- for the 120 hidden variables: <ul> <li>Generates Fig. 9.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_3MRU.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_RefNL_H120.py">SC_MRU/PlotLoss_RefNL_H120.py</a> shows the loss evolution for the different SC-LMSC and SC-MRU-I cells --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- for 120 hidden variables: <ul> <li>Generates Fig. 10.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_RefNL_H120.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_MGRU.py">SC_MRU/PlotLoss_MGRU.py</a> shows the loss evolution for the original and modified MGRU --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- and for the 120 hidden variables: <ul> <li>Generates Fig. A.18.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_MGRU.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_NL.py">SC_MRU/PlotLoss_NL.py</a> shows the loss evolution for the different non-liner transition layers (quadratic and Leaky ReLU) of the SC-MRU-I cell --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- and for the 120 hidden variables: <ul> <li>Generates Fig. B.19.</li> <li><code>Case=1</code> # 0 for quadratic transition blocks and 1 for Leaky ReLU transition blocks</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_NL.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/SC_MRU/PlotLoss_DiffNL_H120.py">SC_MRU/PlotLoss_DiffNL_H120.py</a> shows the loss evolution for the different non-liner transition layers (hybrid) of the SC-MRU-I cell --<a href="#train-the-neural-network-models">after having trained</a> or using saved loss files-- and for the 120 hidden variables: <ul> <li>Generates Fig. B.20.</li> <li>Command to be run from <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> </li> </ul> <div> <pre><code> python3 PlotLoss_DiffNL_H120.py</code></pre> </div> <h2>Part 3: Multiscale analysis</h2> <h3>Trained surrogate models</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/Model">MultiScale/Model</a>: contains the different rnn <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/Bounds_GS">MultiScale/Model/Bounds_GS</a>: Bounds.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/DInpFullModel.pt">MultiScale/Model/DInpFullModel.pt</a>: trained rnn with <code>SC-MRU-T</code> recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/IncrementModel.pt">MultiScale/Model/IncrementModel.pt</a>: trained rnn with <code>SC-MRU-I</code> recurrent cell.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/Model/FullModel.pt">MultiScale/Model/FullModel.pt</a>: trained rnn with <code>SMRU</code> recurrent cell.</li> </ul> </li> </ul> <h3>Run mutiscale simulations using the surrogates</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/model.geo">MultiScale/2D_MultiScale/model.geo</a>: geometry of the macro-scale model that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/model.msh">MultiScale/2D_MultiScale/model.msh</a>: mesh of the macro-scale model that can be read by gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>).</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/model.py">MultiScale/2D_MultiScale/model.py</a>: is used to run all the multiscale simulation: <ul> <li>This requires cm3Libraries (<a href="http://www.ltas-cm3.ulg.ac.be/openSource.htm" target="_blank" rel="nofollow noreferrer noopener">http://www.ltas-cm3.ulg.ac.be/openSource.htm</a>).</li> <li>Uses the bounds and trained models in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/Model">MultiScale/Model</a>.</li> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells <code>cell=["SC_MRU_T","SC_MRU_I","SMRU"]</code>.</li> <li><code>factorStep= 100</code> # is the number of steps x 20 during the reloading stage between points B and C.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/2D_MultiScale">MultiScale/2D_MultiScale/</a> is</li> </ul> </li> </ul> <div> <pre><code>python3 model.py</code></pre> </div> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2">MultiScale/FE2</a>: <ul> <li>Contains the reference displacement-force results of the FE2 simulation.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2/Distributions">MultiScale/FE2/Distributions</a>: contains the macro-scale displacement and stress fields. They can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> </ul> </li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_<code>STEPS</code></a>: <ul> <li>Contain the reference displacement-force results of the rnn-based multiscale simulations for the different recurrent cells <code>CELL</code> and steps number <code>STEPS</code> during the reloading stage between points B and C.</li> <li>For the cases <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_20</a>, the macro-scale displacement and stress fields are also available and can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> </ul> </li> </ul> <h3>Vizualize mutiscale simulations results</h3> <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/2D_MultiScale/plot_force.py">MultiScale/2D_MultiScale/plot_force.py</a>: is used to plot the multiscale simulations results: <ul> <li><code>InpType = cell[0]</code> #choose the recurrent cell typeAmong the different available cells <code>cell=["SC_MRU_T","SC_MRU_I","SMRU"]</code>.</li> <li>The multiscale simulations results to be plotted are saved in the directories <code>CELL_120_Step</code>, where <code>Step</code> # is the number of steps during the reloading stage between points B and C.</li> <li>The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/2D_MultiScale">MultiScale/2D_MultiScale/</a> is</li> </ul> </li> </ul> <div> <pre><code>python3 plot_force.py</code></pre> </div> <ul> <li>To vizualize the macro-scale displacement and stress fields distributions: <ul> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2/Distributions">MultiScale/FE2/Distributions</a>: contains the macro-scale displacement and stress fields. They can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> <li><a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_20</a>, the macro-scale displacement and stress fields are also available and can be vizualized with gmsh (<a href="http://www.gmsh.info" target="_blank" rel="nofollow noreferrer noopener">www.gmsh.info</a>) using the mesh file <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/blob/main/MultiScale/FE2/Distributions/model.msh">model.msh</a>.</li> </ul> </li> </ul> <h2>Part 4: Reproduce paper<sup><a href="#fn-1">1</a></sup> figures</h2> <ul> <li>Fig. 7: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/TrainingPaths">TrainingPaths</a> is</li> </ul> <div> <pre><code> python3 countPathLength.py '../ConstRVE/PathsExamples'</code></pre> </div> <ul> <li>Fig. 9: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> <div> <pre><code> python3 PlotLoss_HiddenV.py</code></pre> </div> <ul> <li>Fig. 10: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> <div> <pre><code> python3 PlotLoss_3MRU.py</code></pre> </div> <ul> <li>Fig. 11: The command to be run from the diretory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU">SC_MRU/</a> is</li> </ul> <div> <pre><code> python3 PlotLoss_DiffNL_H120.py</code></pre> </div> <ul> <li>Figs. 12 and 13: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 LossVS_insertN.py</code></pre> </div> <ul> <li>Figs. 14, 15 and 16: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a></li> </ul> <div> <pre><code> python3 Plot_GS.py</code></pre> </div> <ul> <li>Figs. 17(b)(c)(d): The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/2D_MultiScale">MultiScale/2D_MultiScale/</a> is, see <a href="#vizualize-mutiscale-simulations-results">details</a>:</li> </ul> <div> <pre><code>python3 plot_force.py</code></pre> </div> <ul> <li>Figs. 18-23: Need gmsh to vizualize the results stored in <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale/FE2/Distributions">MultiScale/FE2/Distributions</a> and <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/MultiScale">MultiScale/S<code>CELL</code>_120_20</a>, see <a href="#vizualize-mutiscale-simulations-results">details</a></li> <li>Fig. A.24: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 PlotLoss_MGRU.py.py</code></pre> </div> <ul> <li>Fig. B.25: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 PlotLoss_NL.py</code></pre> </div> <ul> <li>Fig. B.26: The command to be run from the directory <a href="../didearot/didearotPublic/publicationsData/2024_scmru/-/tree/main/SC_MRU/CheckLoss">SC_MRU/CheckLoss/</a> is, see <a href="#vizualize-loss-evolution-and-testing-results">details</a></li> </ul> <div> <pre><code> python3 PlotLoss_DiffNL_H120.py</code></pre> </div> <h2>Disclaimer</h2> <p>This project has received funding from the European Union’s Horizon Europe Framework Programme under grant agreement No. 101056682 for the project “DIgital DEsign strategies to certify and mAnufacture Robust cOmposite sTructures (DIDEAROT)”. The contents of this publication are the sole responsibility of ULiege and do not necessarily reflect the opinion of the European Union. Neither the European Union nor the granting authority can be held responsible for them.</p> <ol> <li> <p>The work is described in:<br>"<em>Wu, L. and Noels, L. (2024).</em> <strong>Self-consistency Reinforced minimal Gated Recurrent Unit for surrogate modeling of history-dependent non-linear problems: application to history-dependent homogenized response of heterogeneous materials</strong> 424: 116881, <a href="https://doi.org/10.1016/j.cma.2024.116881" target="_blank" rel="nofollow noreferrer noopener">doi: 10.1016/j.cma.2024.116881</a>" which can be downloaded. We would be grateful if you could cite this publication in case you use the files. <a href="#fnref-1-5292">↩</a> <a href="#fnref-1-2">↩<sup>2</sup></a> <a href="#fnref-1-3">↩<sup>3</sup></a></p> </li> </ol>

opencc-by-4.0Jan 2024View details →

Toward a Quality Model for Hybrid Intelligence Teams - Supplementary Material

<p>Supplementary material for paper:</p> <div><span>Dell’Anna D, Murukannaiah PK, Dudzik B, Grossi D, Jonker CM, Oertel C, et al.</span> <span>Toward a Qual</span><span>ity Model for Hybrid Intelligence Teams.</span> <span>In: Proceedings of the 23rd International Conference on </span><span>Autonomous Agents and MultiAgent Systems, AAMAS 2024<br><br></span> <div> </div> </div> <div> <div># GENERAL INFORMATION</div> <div> </div> <div> </div> <div>## Title</div> <div> </div> <div>Supplementary material for paper "Toward a Quality Model for Hybrid Intelligence Teams"</div> <div> </div> <div> </div> <div># FILE OVERVIEW</div> <div> </div> <div>```</div> <div>AAMAS2024_Toward_HI_Quality_Model_Supplementary_Material</div> <div>│   README.md</div> <div>│   Group Discussion - Handout - Essentials # The printed version of the questionnaires (Essentials) and team descriptions handed out to the groups for the group discussion</div> <div>│   Group Discussion - Handout - Enablers # The printed version of the questionnaires (Enablers, except Coaching) and team descriptions handed out to the groups for the group discussion</div> <div>│   Group Discussion - Handout - Processes # The printed version of the questionnaires (Coaching and Key Task Processes) and team descriptions handed out to the groups for the group discussion</div> <div>│   Individual Response - Online Survey # A PDF version of the online survey</div> <div>└───Individual Response - Results               # Excel file with all the responses of the participants from the online survey. Excel sheets are organized as follows</div> <div>RAW DATA                                # all the raw data from the survey</div> <div>RAW DATA TRANSPOSED                     # raw data with rows and columns inverted</div> <div>ALL TEAMS SCORES                        # quantitative results about adequacy for each feature: team scores and importance</div> <div>ALL FEEDBACK PER FEATURE                # qualitative results about adequacy for each feature: grouped feedback participants</div> <div>forlatex                                # results in latex-ready form</div> <div>forspider                               # results in a form that supports easy creation of spider plot</div> <div>EFFECTIVENESS                           # quantitative results about effectiveness measures</div> <div>FINAL COMMENTS                          # final qualitative feedback from participants</div> <div>essentials                              # RAW DATA TRANSPOSED (essentials only)</div> <div>enablers                                # RAW DATA TRANSPOSED (enablers only)</div> <div>processes                               # RAW DATA TRANSPOSED (processes only)</div> <div>*_num_scores                            # * in {essentials, enablers, processes}, likert-to-numerical scores</div> <div>*_num_scores_annotatio                  # same as *_num_scores, just with numerical values instead of formulas</div> <div>```</div> <div> </div> <div> </div> <div> </div> <div> </div> </div>