Find research datasets worth reusing

opencc-by-4.0Aug 2018View details →

Trained neural network data for synchrotron radiative transfer in the Stokes basis, power law model, computed by rimphony, for consumption by neurosynchro

This archive contains data representing a trained-up neural network suitable for use with the <a href="https://github.com/pkgw/neurosynchro/">neurosynchro</a> package. The network generates coefficients that can be used for numerical radiative transfer of synchrotron emission in the Stokes basis with a package such as <a href="https://github.com/jadexter/grtrans/">grtrans</a>. In this particular dataset, networks were trained on a training set of coefficients generated by <a href="https://github.com/pkgw/rimphony/">rimphony</a> that is available as <a href="https://doi.org/10.5281/zenodo.1341154">DOI:10.5281/zenodo.1341154</a>. The data were generated using a model of a power law electron distribution isotropic in pitch angle. The input parameters, which were sampled randomly in a three-dimensional space, were: <ul> <li>s, the harmonic number, dimensionless, sampled logarithmically between 5 and 50,000,000.</li> <li>theta, the angle between the ray path and the local magnetic field, measured in radians, sampled linearly between 0.001 and π/2 (namely, 1.5707963267948966).</li> <li>p, the power-law index of the energetic electrons, dimensionless, sampled linearly between 1.5 and 7.</li> </ul> The training set was computed on Harvard’s Odyssey cluster using Git commit <a href="https://github.com/pkgw/rimphony/commit/772161ebda0217b8c1ccb8ce3801ad9dc3701a4f">772161</a> of rimphony. A total of about 5,000 CPU hours were used, with 500 processes running for about 10 hours each, yielding about 22 million numbers. Training the networks took about 3 hours on an 8-core laptop. For the purposes of neurosynchro, the formats of the files in this package should be regarded as internal implementation details. The <a href="https://pypi.org/project/neurosynchro/">neurosynchro</a> Python package will load up the files in this archive and use them to predict synchrotron coefficients. For specifics, see <a href="https://neurosynchro.readthedocs.io/en/stable/">the neurosynchro documentation</a>.

opencc-by-nc-4.0Oct 2018View details →

gazeNet: End-to-end eye-movement event detection with deep neural networks

This repository contains synthetic eye-movement dataset used to train deep learning based eye-movement event detection algorithm described in Zemblys, R., Niehorster, D.C. & Holmqvist, K. (2018). gazeNet: End-to-end eye-movement event detection with deep neural networks. Behavior research methods, pp 1–25. <a href="https://doi.org/10.3758/s13428-018-1133-5">https://doi.org/10.3758/s13428-018-1133-5</a> Code used to train a model can be found on github <a href="https://github.com/r-zemblys/gazeNet">here</a>. Code to generate synthetic eye-movement data can be downloaded from <a href="https://github.com/r-zemblys/gazeGenNet">here</a>.

opencc-by-4.0Nov 2018View details →

Revealing Ferroelectric Switching Character Using Deep Recurrent Neural Networks

The ability to manipulate domains and domain walls underpins function in a range of next-generation applications of ferroelectrics. While there have been demonstrations of controlled nanoscale manipulation of domain structures to drive emergent properties, such approaches lack an internal feedback loop required for automation. Here, using a deep sequence-to-sequence autoencoder we automate the extraction of features of nanoscale ferroelectric switching from multichannel hyperspectral band-excitation piezoresponse force microscopy of tensile-strained PbZr0.2Ti0.8O3 with a hierarchical domain structure. Using this approach, we identify characteristic behavior in the piezoresponse and cantilever resonance hysteresis loops, which allows for the classification and quantification of nanoscale-switching mechanisms. Specifically, we are able to identify elastic hardening events which are associated with the nucleation and growth of charged domain walls. This work demonstrates the efficacy of unsupervised neural networks in learning features of the physical response of a material from nanoscale multichannel hyperspectral imagery and provides new capabilities in leveraging multimodal in operando spectroscopies and automated control for the manipulation of nanoscale structures in materials.

opencc-by-4.0Jan 2019View details →

Domain-Independent Reviews' Sentiment Polarity Classification using Shallow Word2Seq Convolutional Neural Network

Reviews and comments are perceptions about specific services or products. They are embedded with hidden sentiments the reviewer has towards certain subjects. Business owners use customer reviews to understand customers’ perceptions about specific services or products. The ability to understand reviews’ sentiment from different domains or areas give decision makers and business owners the opportunity to make critical business decisions which can help them to increase profits of their businesses. Previous studies had focussed on classifying sentiment polarity by using traditional machine learning and deep learning methods. However, these suffered from low model generalization ability, causing the models to perform better only on single domain datasets rather than multiple domain datasets. The problem is the inability of the classification model to learn domain-restricted knowledge from multi-domain datasets. Aiming to improve the accuracy of the cross-domain classification, this paper proposes a method which uses Word2Seq Convolutional Neural Network (CNN) to classify reviews’ sentiment across multiple domain datasets (i.e. digital worker, movie, product, hotel and restaurant reviews). The evaluation showed that the proposed method had achieved the state-of-the-art performance. The high classification performance also promoted the reliability and effectiveness of implementing the Word2Seq CNN to classify reviews’ sentiment across different domains and learn domain restricted knowledge while improving the model generalization ability. The uploaded dataset is a sampled dataset with 5000 observations for both training and testing sets.

opencc-by-4.0Sep 2019View details →

A Convolutional Neural Network classifier identifies tree species in mixed-conifer forest from hyperspectral imagery

Published online: <a href="https://www.mdpi.com/2072-4292/11/19/2326">https://www.mdpi.com/2072-4292/11/19/2326</a> DOI: 10.3390/rs11192326 Abstract: In this study, we automate tree species classification and mapping using field-based training data, high spatial resolution airborne hyperspectral imagery, and a convolutional neural network classifier (CNN). We tested our methods by identifying seven dominant trees species as well as dead standing trees in a mixed-conifer forest in the Southern Sierra Nevada Mountains, CA (USA) using training, validation, and testing datasets composed of spatially-explicit transects and plots sampled across a single strip of imaging spectroscopy. We also used a three-band ‘Red-Green-Blue’ pseudo true-color subset of the hyperspectral imagery strip to test the classification accuracy of a CNN model without the additional non-visible spectral data provided in the hyperspectral imagery. Our classifier is pixel-based rather than object based, although we use three-dimensional structural information from airborne Light Detection and Ranging (LiDAR) to identify trees (points > 5 m above the ground) and the classifier was applied to image pixels that were thus identified as tree crowns. By training a CNN classifier using field data and hyperspectral imagery, we were able to accurately identify tree species and predict their distribution, as well as the distribution of tree mortality, across the landscape. Using a window size of 15 pixels and eight hidden convolutional layers, a CNN model classified the correct species of 713 individual trees from hyperspectral imagery with an average F-score of 0.87 and F-scores ranging from 0.67–0.95 depending on species. The CNN classification model performance increased from a combined F-score of 0.64 for the Red-Green-Blue model to a combined F-score of 0.87 for the hyperspectral model. The hyperspectral CNN model captures the species composition changes across ~700 meters (1935 to 2630 m) of elevation from a lower-elevation mixed oak conifer forest to a higher-elevation fir-dominated coniferous forest. High resolution tree species maps can support forest ecosystem monitoring and management, and identifying dead trees aids landscape assessment of forest mortality resulting from drought, insects and pathogens. We publicly provide our code to apply deep learning classifiers to tree species identification from geospatial imagery and field training data Digital Publication of the training data polygons and hyperspectral imagery used in the manuscript "A Convolutional Neural Network classifier identifies tree species in mixed-conifer forest from hyperspectral imagery". Code is available in a Jupyter Notebook and can be found here: <a href="https://github.com/jonathanventura/canopy">https://github.com/jonathanventura/canopy</a> National Ecological Observatory Network. 2018. Provisional data downloaded from <a href="http://data.neonscience.org/">http://data.neonscience.org</a> on 22 June 2018. Battelle, Boulder, CO, USA

opencc-by-4.0Dec 2008View details →

Fig. 3 in MAX WEBER AN ARTIFICIAL SOCIOLOGY. PROPOSAL FOR A NEURAL NETWORK WITH WEBERIAN REASONING

Fig. 3. The ornithopod dinosaur Gasparinisaura cincosaltensis Coria and Salgado, 1996 from the Late Cretaceous Anacleto Formation of Patagonia, MCSPv 111 (A) and MCSPv 112 (B). A1, nearly complete postcranial skeleton in left lateral view; A2, cluster of gastroliths in the abdominal cavity; B1, nearly complete skeleton in dorsal view; B2, the largest cluster of gastroliths below the last dorsal vertebra.

opencc-by-4.0Dec 2008View details →

Fig. 2 in MAX WEBER AN ARTIFICIAL SOCIOLOGY. PROPOSAL FOR A NEURAL NETWORK WITH WEBERIAN REASONING

Fig. 2. The ornithopod dinosaur Gasparinisaura cincosaltensis Coria and Salgado, 1996, MUCPv 213 from the Late Cretaceous Anacleto Formation of Patagonia. A. Forelimb bones and ribs. B. The longest gastrolith in contact with two right dorsal ribs. C. Cluster of gastroliths associated with the ribs. D. Scanning electron microphotograph of a gastrolith from metamorphic rock.

opencc-by-4.0Dec 2008View details →

Fig. 1 in MAX WEBER AN ARTIFICIAL SOCIOLOGY. PROPOSAL FOR A NEURAL NETWORK WITH WEBERIAN REASONING

Fig. 1. The surroundings of Cinco Saltos City where MCSPv 111, MCSPV 112 (Site 1) and MUCPv 213 (Site 2) were found (modified from Andreis et al. 1974).

opencc-zeroJul 2024View details →

Deep Learning Neural Network Development for the Classification of Bacteriocin Sequences Produced by Lactic Acid Bacteria

This project contains the following underlying data: <h3>      Software-Related Files</h3>                ·         BacLABNet_script.ipynb (Deep Learning Neural Network for classification of Bacteriocin Sequences)                 ·         embed_proteins.py (Recurrent Neural Network to obtained the embedding vectors)                ·         model_I22.h5 (This file contains the trained weights of the trained model)                ·         model_I22.json (This file contains the structure of the trained model)                ·         rnn_gru.pt (Initial weights of the Recurrent Neural Network to obtain embedding vectors)                ·         List_kmers.csv (List of 5-mers and 7-mers obtained from dataset after it filtered sequences shorter than 50 aa and                           longer than 2000 aa)                     Files Used for Training, Testing, and Validation of the Neural Network                ·         data_nonBacLAB.csv (25000 nonBacLAB amino acid sequences retrieved from Uniprot)                ·         data_BacLAB.csv (24964 BacLAB amino acid sequences retrieved from Uniprot)              Additional Files                 ·         data_BacLAB_and_nonBacLAB.csv (Combination of sequences from data_BacLAB.csv and data_nonBacLAB.csv)                  ·         all k.mers list.xlsx (Table of all k-mers obtained for k=3,5,7,15,20)   Note: Codes are additionally available on GitHub.  Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication)(http://creativecommons.org/publicdomain/zero/1.0/)

Results of Improved SMAP Soil Moisture Retrieval Using a Deep Neural Network-based Replacement of Radiative Transfer and Roughness Model

This repository contains: <ol> <li>A deep neural network (DNN) based soil moisture (SM) estimates (NN) based on the SMAP TB (Descending, 6 AM) and SMAP SCA-V ancillary data as the input variables. (<a href="../api/records/13309165/draft/files/SMAP_NN_36km_20150331_20220326.nc/content" target="_blank" rel="noopener noreferrer">SMAP_NN_36km_20150331_20220326.nc</a>)</li> <li>Temporally averaged roughness parameter (hNN) and scattering albedo (omegaNN) which are retrieved by inversely tracking the DNN model. (<a href="../api/records/13309165/draft/files/SMAP_hNN_omegaNN_36km_temporal_average_201503_202103.nc/content" target="_blank" rel="noopener noreferrer">SMAP_hNN_omegaNN_36km_temporal_average_201503_202103.nc</a>)</li> </ol> Summary: The DNN model has been developed by relating SMAP TB and SMAP SCA-V ancillary data with in-situ SM data from the international soil moisture network (ISMN) using DNN. To minimize scale mismatch between gridded SMAP data and point in-situ data, the triple collocation analysis was conducted. The SM estimated from the DNN algorithm (NN) showed a good agreement with the ISMN data that was not used in the model training. Moreover, for a densely vegetated region located in the Amazon (Tambopata site) the NN showed less bias compared to available SM retrievals.  Two parameters hNN and omegaNN are retrieved by ingesting NN to the modified dual channel algorithm. When the SM retrieval was conducted using the hNN and omegaNN, the result showed good agreement with the NN (DNN-based SM) with R of 0.986, ubRMSD of 0.015 m3/m3, and bias of -0.001 m3/m3. The paper "Improved SMAP Soil Moisture Retrieval Using a Deep Neural Network-based Replacement of Radiative Transfer and Roughness Model" published in the Transactions on Geoscience and Remote Sensing. For more details, please contact me (wotp12@unist.ac.kr)

SDUST2023BCO: a global seafloor model determined from multi-layer perceptron neural network using multi-source differential marine geodetic data

<div> SDUST2023BCO.nc is the global marine bathymetric model covering 80°S~80°N and 0°~360°E on 1′×1′ grids. The dataset contains geospatial information (latitude, longitude), SDUST2023BCO bathymetric model and an attachment data. </div>

Assets (code, scripts and datasets) for the manuscript "Correction of the Air-Sea Heat Fluxes in Ocean General Circulation Models Using Neural Networks"

This dataset contains all relevant software and data related to the manuscript "Correction of the Air-Sea Heat Fluxes in Ocean General Circulation Models Using Neural Networks", submitted to AGU journals.

opencc-by-4.0May 2023View details →

Data archive for 'Convolutional neural networks facilitate river barrier detection and evidence severe habitat fragmentation in the Mekong River biodiversity hotspot'

This repository contains the code and databases used in the paper 'Convolutional neural networks facilitate river barrier detection and evidence severe habitat fragmentation in the Mekong River biodiversity hotspot'.  The 'Mekong River Barrier Database (MRBD)' folder contains the basin-scale barrier database developed in this study. This database contains more than 13,000 unique barriers, which were identified by using the convolutional neural networks-based object detection method from Google Earth’s satellite imagery. The 'FCOS' folder contains the barrier detection model (FCOS ResNext-101-FPN), trained for detecting river barriers from remotely sensed images within the MMDetection framework. The 'FCOS_x101_v2' folder contains the enhanced FCOS model. The 'R_script' folder contains R files used in the paper. Coordinate.R was used to extract coordinates from bounding boxes in each TIF image. CAFI.R was used to calculate the CAFI index in each sub-catchment. The 'Barrier image training set' folder contains over 10,000 river barrier satellite images and their associated JSON files, forming the 'training, validation, and test datasets' used during the model training process. This dataset is made available to the user community in raw, in the hope that others will contribute to its future development, thereby enhancing its use and utility. For more information on the MMDetection framework, refer to the following GitHub repository: <a href="https://github.com/open-mmlab/mmdetection">https://github.com/open-mmlab/mmdetection</a>

Neural-network-based molecular dynamics simulations reveal that proton transport in water is doubly gated by sequential hydrogen-bond exchange: Neural network potentials training data

<h1>Neural network potentials of an excess proton in bulk water, training data</h1> This dataset contains 2188 configurations labeled at two hybrid DFT levels (revPBE0-D3 and B3LYP-D3). The configurations are given as a single XYZ file: configurations.xyz The box dimensions are written in box.txt The energies for all configurations at a given level of theory are written in energies_LEVEL.txt (one configuration per line) The atomic forces for each configuration at a given level of theory are gathered in a XYZ file: forces_LEVEL.xyz The relative displacements of the Wannier centroids, with respect to the closest oxygen atom, for each configuration at a given level of theory, are in the following XYZ file: wannier-centroids-displacements_LEVEL.xyz

Training data set for: Graph Neural Network based elastic deformation emulators for magmatic reservoirs of complex geometries

<h2>Overview</h2> This is a synthetic volcano deformation dataset accompanying the publication of Graph Neural Network based elastic deformation emulators for magmatic reservoirs of complex geometries, on the journal Volcanica. Synthetic, quasi-static deformation is computed for magma chambers of various geometries, parameterized as spheroids or superpositions of spherical harmonics. Surface deformation is computed using the boundary element method (BEM) of Nikkhoo & Walter (2015). Please reference our paper for details of computational methods. The dataset contains 50,000 realizations of magma chamber geometries/orientations/centroid depths and associated deformation fields. Surface deformation fields are sampled at discrete locations, with a uniform random distribution within [Lh x Lh], and a distribution that concentrates near the chamber (at radial distances, r = 10^(-3  random number) * Lh/2). Note this dataset contains only a small fraction of the total dataset. In total, 824,393 realizations of magma chambers were used to train our emulators. For accessing the complete training data set, please contact the authors.  Each .mat file contains the deformation field associated with a single chamber geometry. Use visData.m to visualize chamber geometry and associated surface displacement. Each file contains two MATLAB structures, "input" and "output".  <h2>Naming of each zip file</h2> The numbers after the underscore, N:M, indicate that this file contains N of the M total chamber realizations for this particular setup.  <a href="../api/records/13800065/draft/files/sph_20AspRatios_1e4:151211.zip.zip/content" target="_blank" rel="noopener noreferrer">sph_20AspRatios_1e4:151211.zip</a>: deformation corresponding to spheroidal magma chambers parameterized by aspect ratios.  <a href="../api/records/13800065/draft/files/sh_complex_1e4:152283.zip/content" target="_blank" rel="noopener noreferrer">sh_complex_1e4:152283.zip</a>: deformation corresponding to chamber geometry produced by superposition of spherical harmonic modes.  <a href="../api/records/13800065/draft/files/sh_mode_approx_1e4:138380.zip/content" target="_blank" rel="noopener noreferrer">sh_mode_approx_1e4:138380.zip</a>: deformation corresponding to chamber geometries corresponding to individual spherical harmonic modes, combined with a spherical mode (the spherical mode prevents chamber surfaces from having zero radii locally) <a href="../api/records/13800065/draft/files/sh_spheroid_approx1e4:202272.zip/content" target="_blank" rel="noopener noreferrer">sh_spheroid_approx1e4:202272.zip</a>: deformation corresponding to chambers approximating spheroids, but parameterized by spherical harmonics. <a href="../api/records/13800065/draft/files/sh_spheroid_perturb_1e4:180247.zip/content" target="_blank" rel="noopener noreferrer">sh_spheroid_perturb_1e4:180247.zip</a>: same as above, but with additional random perturbations parameterized in spherical harmonics. <h2>Variables in each file</h2> Input contains the following fields: dp2mu: pressure change to shear modulus ratio. dx, dy, dz: the coordinates of chamber centroid [meters] mu: dimensionless crustal shear modulus (always set to 1) nu: crustal Poisson's ratio (always set to 0.25) Ns: number of points on the surface where displacements are computed Lh, Lv: horizontal and vertical dimensions of the model domain [meters]. Lh is determined such that at the edge of the model domain, the displacement magnitude is below 10 percent of the maximum. Lv = Lh/2 + abs(dz) for the spheroids ----------------------------------------------------------------------------------------------------------- the input files contain asp: aspect ratio of chamber (length of the semi-major axis divided by that of the semi-minor axis) ra, rb: semi-major, -minor, axis length [meters] thetax, thetay, thetaz: counterclockwise rotation angles with regard to x, y, z axis [degrees]. thetax = [0, 90] degrees, thetay = 0 degrees, thetaz = 360 degrees. for the general geometries-------------------------------------------------------------------------------------------------- the input files contain ls, ms, fs: degree, order, coefficients of spherical harmonic modes. Spherical harmonics are sampled up to degree 5. fs is a complex vector of coefficients such that the resulting shape is real.  normF: normalization factor applied to the shape parameterized by ls, ms, fs, such that the shape as a maximum radius of unity. rmax: scale factor to scale the spherical harmonics parameterized shape to real dimensions [meters]. ============================================================================================= Output contains the following fields, X, Y, Z: coordinates of points where displacement vectors are computed [meters] Ux, Uy, Uz: displacements in x, y, z directions [meters] P, T: coordinates [meters] of vertices for the triangular mesh used in BEM calculation, and the connectivity matrix  C: coordinates [meters] of the center of each triangular element that, dhat, nhat: unit vectors for orthogonal coordinate systems local to each triangular element. that ("t-hat") extends from vertex one to vertex two, nhat is outward normal, and dhat = cross (nhat, that). Reference: 1. Nikkhoo, M., & Walter, T. R. (2015). Triangular dislocation: an analytical, artefact-free solution. Geophysical Journal International, 201(2), 1119-1141.

Data and models for: Learning Ordering in Crystalline Materials with Symmetry-Aware Graph Neural Networks

Data (ver 1.1) and trained models for our paper "<a href="https://arxiv.org/abs/2409.13851">Learning Ordering in Crystalline Materials with Symmetry-Aware Graph Neural Networks</a>". If you use such data or models, please cite our paper. These three directories need to be downloaded and copied into our source codes in order to reproduce our paper: <a href="https://github.com/learningmatter-mit/PerovskiteOrderingGCNNs">https://github.com/learningmatter-mit/PerovskiteOrderingGCNNs</a> <ul> <li>data: All data files for training and evaluating GCNNs, with a copy archived on the Materials Data Facility (<a href="https://doi.org/10.18126/ncqt-rh18">DOI: 10.18126/ncqt-rh18</a>)</li> <li>saved_models: All saved model files for evaluating GCNNs</li> <li>best_models: All best model files for evaluating GCNNs</li> </ul>

Data for "Using physics-informed neural networks to predict the lifetime of laser powder bed fusion processed 316L stainless steel under multiaxial low-cycle fatigue loading"

Title of dataset: Data for "Using physics-informed neural networks to predict the lifetime of laser powder bed fusion processed 316L stainless steel under multiaxial low-cycle fatigue loading". Name/institution/contact information: Dr. Michal Barto&scaron;ák, Czech Technical University in Prague - Faculty of Mechanical Engineering, email: michal.bartosak@fs.cvut.cz. Date of data collection: The data were collected between 2021 and 2024. File name structure: The data consists of two files: "316L_fatigue_and_defects.xls," which contains fatigue lifetime data and defect characteristics, and an associated description file, "read_me.txt." See "https://doi.org/10.1016/j.ijfatigue.2024.108608" for the associated article and a detailed description of the methods.

Datasets used in "Assesing the quality of random number generators through neural networks", Machine Learning: Science and Technology 5 (2024) 025072

Datasets corresponding to the bits generated by different random number generators used in J. L. Crespo et al, Machine Learning: Science and Technology 5 (2024) 025072. VCSEL_QRNG_postprocessed_bits.txt: postprocessed bits from the random generator based on gain-switching of VCSELs  EC_LCG_bits.txt:  bits from the linear congruential generator on elliptic curves LCG_32_bits.txt:: bits from the linear congruential generator with 32 bits VCSEL_QRNG_raw_bits.txt: raw bits from the random generator based on gain-switching of VCSELs

A combined NMR and Deep Neural Network approach for enhancing the spectral resolution of aromatic side chains in proteins

Nuclear magnetic resonance (NMR) spectroscopy has become an important technique in structural biology for characterising the structure, dynamics and interactions of macromolecules. While a plethora of NMR methods are now available to inform on backbone and methyl-bearing side-chains of proteins, a characterisation of aromatic side chains is more challenging and often requires specific labelling or 13C-detection. Here we present a deep neural network (DNN) named FID-Net-2, which transforms NMR spectra recorded on simple uniformly 13C labelled samples to yield high-quality 1H-13C correlation spectra of the aromatic side chains. Key to the success of the DNN is the design of a complementary set of NMR experiments that produce spectra with unique features to aid the DNN produce high-resolution aromatic 1H-13C correlation spectra with accurate intensities. The reconstructed spectra can be used for quantitative purposes as FID-Net-2 predicts uncertainties in the resulting spectra. We have validated the new methodology experimentally on protein samples ranging from 7 to 40 kDa in size. We demonstrate that the method can accurately reconstruct high resolution two-dimensional aromatic 1H-13C correlation maps, high resolution three-dimensional aromatic-methyl NOESY spectra to facilitate aromatic 1H-13C assignments, and that the intensities of peaks from the reconstructed aromatic 1H-13C correlation maps can be used to quantitatively characterise the kinetics of protein folding.  More generally, we believe that this strategy of devising new NMR experiments specifically for analysis using customised DNNs represents a substantial advance that will have a major impact on the study of molecules using NMR in the years to come.