Articles tagged with Data Sets
Researchers at Dartmouth College used the Bible to develop an algorithm that can convert written works into different styles for different audiences. The study, published in Royal Society Open Science, trained on over 1.5 million unique pairings of source and target verses from various versions of the sacred texts.
A study led by OHSU has published the largest cancer dataset of its kind, involving hundreds of patient samples and revealing new insights into acute myeloid leukemia (AML). The dataset may help researchers and physicians solve specific questions about AML treatment options.
Two types of flash drought have been identified in China, Type I driven by high temperatures and evapotranspiration, while Type II is initiated by rainfall deficiency. Both types have increased significantly over the past 30 years, with a two-fold greater increase for Type I.
Researchers at the University of Sussex have created a large dataset that can be used to develop smartphone apps that detect transportation modes, predict road conditions, and offer route recommendations. The dataset has achieved high accuracy rates in recognizing modes of transport, paving the way for innovative mobile applications
A deep learning approach can identify nuanced mammographic imaging features specific to recalled but benign (false-positive) mammograms, distinguishing them from those identified as malignant or negative. The study achieved an area under the curve (AUC) of 0.76-0.91, indicating high performance in detecting false recalls.
The new global marine environmental forecasting system will provide accurate predictions of ocean temperature, salinity, velocity, and tidal currents with a horizontal resolution of 10 km. The system's accuracy is expected to reach an advanced international level, enabling clear identification of ocean mesoscale phenomena.
A new MIT model has been developed to reduce false positives in credit card fraud detection, achieving a 54% reduction in incorrect flagging of legitimate transactions. By extracting more than 200 detailed features for each individual transaction, the model can better pinpoint unusual spending habits and improve accuracy.
Researchers at Northwestern University have identified four clusters of personality types using comprehensive data from over 1.5 million questionnaire respondents. The findings challenge existing paradigms in psychology and provide new insights into human behavior.
Researchers at MIT develop Temporal Relation Network (TRN) module to help CNNs recognize activities by observing key frames. The module achieves top accuracy of 95% in activity recognition on Jester dataset, outperforming existing models.
Researchers found that overconfident CEOs are 33% more likely to be sued by shareholders. A shareholder lawsuit can curb future risk-taking behavior, leading to lower confidence and more prudent actions. This study highlights the importance of shareholder power in regulating CEO behavior.
A Lehigh research team is developing a 'Google for research data' that can assist scientists in locating datasets across various disciplines. The team aims to create a domain-agnostic search engine that utilizes user-centric methods to develop dataset search tools and novel indexing techniques.
Researchers from the Santa Fe Institute developed a new algorithm called SpringRank that analyzes wins and losses in networks to predict outcomes. The algorithm outperformed others in efficiency and accuracy, even when applied to diverse datasets such as NCAA basketball teams and animal social behaviors.
The DeepLesion database is the largest publicly available medical image dataset, containing over 32,000 annotated lesions from 10,000 case studies. It has tremendous potential to jump-start the field of computer-aided detection and diagnosis.
A Massachusetts-based study found that opioid overdose events decreased during pregnancy but increased in the postpartum period, particularly when fewer resources were available to mothers with substance use disorder. The research team developed a unique dataset linking statewide resources and found that women with opioid use disorder ...
A new approach uses machine learning to generate computer-generated X-rays to supplement real images, increasing the size of training sets for AI systems. This method improves classification accuracy for common and rare conditions by up to 40%, overcoming a challenge in applying artificial intelligence to medicine.
A new study by Queen Mary University of London found that 58% of online video piracy is concentrated in just two locations, making them susceptible to detection by copyright enforcers. The research also revealed a global network of streaming cyberlockers and third-party indexing services facilitating piracy.
The GA4GH Large Scale Genomics Work Stream introduces the htsget protocol, a standard for accessing large-scale genomic sequencing data online without file transfers. This enables global sharing and collaboration, addressing the growing need for big-data cloud-based approaches in genomics.
A new study by North Carolina State University researchers found that companies incorporated in tax haven countries with weak governance are more likely to engage in practices benefiting executives at the cost of their shareholders. These companies paid an average of 83% less in dividends to shareholders, compared to those in well-regu...
Researchers developed a robust model for general causality that identifies multiple causal connections without time-sequence data. The Multivariate Additive Noise Model (MANM) works on simulated real-world datasets, offering opportunities to analyze complex phenomena in areas like economics and disease outbreaks.
Researchers developed a computational framework to analyze large-scale single-cell gene expression levels, enabling the study of unprecedented cellular heterogeneity in rare cell populations. The BigSCale tool successfully processed 1.3 million individual cells from a mouse brain dataset.
A new study by the University of Waterloo confirms that mergers create value for investors, improving shareholder value and market share. The research found that firms realize synergies from mergers, benefiting all stakeholders, including consumers who do not face higher prices.
Researchers at UT Dallas developed a tool to identify inconsistencies in cosmological data, revealing potential errors in current models. The findings suggest that either systematic errors need to be removed or the underlying model is incomplete, leading to questions about Albert Einstein's theory of gravity.
The NIMS inorganic material database AtomWork-Adv has been made available to the public with enhanced features. The database contains a vast amount of crystal structure, X-ray diffraction, material properties, and phase diagram data collected from literature up to 2014.
A study found a significant link between low birth weight and left-handedness in two large sets of Japanese and Dutch newborn children. Left-handers averaged lower birth weights, with differences observed even within families.
The USGS and DOE have released a comprehensive dataset of U.S. wind turbine locations and characteristics, enabling accurate planning and research. The database contains over 57,000 turbines and allows users to search and sort the wind fleet by various criteria.
Researchers used synthetic-aperture radar data from four satellites to analyze the Lake Urmia Causeway in Iran, finding accelerated deformation due to soil consolidation and human activity. They also developed a predictive model for future deformation, highlighting the potential of space-based monitoring for critical structures.
Researchers developed a new method to train computers to better recognize objects in the real world by using virtual reality. A virtual dataset called ParallelEye was created, allowing for diverse and realistic images of various scenes, which significantly improved performance on object detection tasks.
Researchers analyzed 136 million keystrokes from 168,000 volunteers to identify what makes a faster typist. The study found that the fastest typists use rollover typing, where they press the next key before the previous one is released, and display different typing styles.
A team of computational biologists developed an algorithm to integrate multiple sequencing datasets at single-cell resolution. This approach enables the comparison of single-cell datasets and dissects differences between them.
Researchers have created a massive 13 million-person family tree using genealogy data, revealing trends in marriage, migration, and longevity. The study found that people are more likely to marry fourth cousins than seventh cousins, and that women in Europe and North America have migrated more than men.
A new study finds that over 99% of Chicagoland wetlands are home to non-native plant species, with reed canary grass dominating many sites. The research highlights the importance of considering ecosystem function and functional values in restoration efforts in urban areas.
The Clinical Epidemiology Database aims to facilitate collaboration among researchers by providing a standardized platform for accessing and exploring complex clinical data. The database, launched by the University of Pennsylvania, introduces an intuitive interface and provides documentation of study design and background.
The USC-led ATLAS project has compiled a large dataset of brain scans from stroke patients, which is now available for download. Researchers are using this dataset to develop and test algorithms that can automatically process MRI images from stroke patients, aiming to identify biological markers for personalized treatment plans.
LMIC researchers face concerns over national ownership and unequal publishing opportunities when sharing open datasets. To address these issues, increased investment in LMIC data management infrastructure and international policy frameworks are proposed.
Researchers have developed a new method that can accurately estimate biodiversity at a global scale, up to 10 orders of magnitude. The study used the 1999 Great Britain Countryside Survey dataset and tested over a dozen methods, with one model providing estimates within 10% of the true value.
Researchers used a combination of conventional and remotely-sensed climate measurements to model the future distributions of bamboo species in southwestern China. This approach increased confidence in model results, providing valuable guidance for conservation planning and nuanced decision-making.
Researchers developed codes MENNDL and RAVENNA to efficiently design and train neural networks, generating and training up to 18,600 networks simultaneously. This enables the training of highly accurate networks in a fraction of the time, with applications in self-driving cars, intelligent robots, and scientific experiments.
The CARICOMP program found significant decreases in water quality at 42% of the monitoring stations across the Caribbean basin. Despite expected increases in water temperature due to global warming, no such changes were detected in the data set.
A computational study by Queen Mary University of London has identified Botswana as the country with the most distinct musical recordings globally. The researchers found that China has the most spatial outliers compared to its neighboring countries, highlighting unique timbral characteristics such as the butterfly harp string instrument.
Scientists at Vrije Universiteit Amsterdam found that averaging all estimates yields significant accuracy gains, especially when considering multiple people. Taking the average of estimates from different individuals is a more effective approach for good decision-making than relying on personal expertise.
Scientists analyze NASA data to investigate solar storm impact on cetaceans' internal compasses. While space weather is not primary driver of strandings, it may be one factor among several contributing to the phenomenon.
Researchers developed a computational tool that combines cellphone records with satellite and geographic information systems to create timely and incredibly detailed poverty maps. The study focused on Senegal, creating maps detailing the poverty levels of 552 communities.
A new study by researchers at LMU Munich confirms that sponges are the oldest animal group, contradicting previous findings that suggested comb jellies were the first. The team used a refined method to analyze genetic data and found that sponges' simple multicellular structure is more primitive than initially thought.
Recent genomic analyses have flipped between whether sponges or comb jellies are the oldest lineage of living animals. New research by Professor Davide Pisani reveals that sponges are the most ancient lineage, with models that describe the data poorly favouring comb jellies and those that better describe the data favouring sponges.
A new study finds that in-store product sampling has both immediate and sustained effects on sales, with smaller stores benefiting more. The model also reveals that repeated sampling for a single product increases returns and expands product categories.
Researchers efficiently used Stampede2's 1024 Skylake processors to complete a 100-epoch ImageNet training with AlexNet in 11 minutes, setting the fastest time recorded to date. The Layer-Wise Adaptive Rate Scaling (LARS) algorithm enabled this breakthrough, allowing for larger-than-ever batch sizes and adaptive learning rate adjustments.
The Illinois Waterway has successfully recovered following the passage of the 1972 Clean Water Act, according to a study published in BioScience. The authors used a robust data set to demonstrate a strong linkage between water quality and the rebound of fish populations.
Scientists have now shown that individual neurons in the mouse brain often spool out spindly fibers nearly half a meter long. The team's dataset and analysis tool, MouseLight NeuronBrowser, offer new insights into how elaborately-branching neurons route information in the brain.
The Energy efficiency Data Portal & Protocol (EeDaPP) Initiative aims to create a standardized reporting template for loan-by-loan energy efficient mortgage assets. This will enable large-scale tracking of performance and facilitate earmarking for energy efficient bond issuance.
Scientists have unveiled a dataset mapping diverse migrant communities in the EU, providing insights for policymakers to develop targeted integration policies. The maps reveal residential patterns of migrants by nationality or country of origin, highlighting concentration, diversity, and segregation within cities.
Researchers developed a method to extract Higgs boson signal from noise data using quantum-compatible machine learning techniques, outperforming standard counterparts even with small datasets. The new approach is expected to be useful for problems beyond high-energy physics.
The Critical Assessment of Metagenome Interpretation (CAMI) Challenge evaluated computational tools for metagenomes, assessing assemblers, binners, and taxonomic profilers. The benchmarking results provide performance overviews for developers and applied scientists, informing the selection of suitable software for research questions.
The university's researchers compiled a large solar dataset from NASA's Solar Dynamics Observatory mission, making several hundred thousand solar events available to the public. The dataset has improved the quality of the data and will accelerate computer vision research on these solar images.
Researchers found a lower density in Mars' crust, indicating possible porosity, which could impact the planet's formation and evolution. The new gravity map reveals variations in crust thickness beneath volcanoes.
A recent study published in BMC Ecology found that grey mouse lemurs with high hair cortisol levels had a lower chance of survival. The researchers discovered that low cortisol levels were associated with a 13.9% higher chance of survival compared to those with high levels.
The study examined three types of distracting elements: sensorimotor, cognitive, and emotional. Researchers found that texting led to more dangerous driving, while a 'sixth sense' protected drivers from emotional upset or absent-mindedness.
Researchers from NIST and Michigan State University have developed an algorithm that automates the key decision point in fingerprint analysis, reducing human subjectivity and improving reliability. The new system uses machine learning to score latent prints based on their quality, allowing for more efficient processing of evidence.
The study found that brain activity in regions called the default mode network represent a kind of higher-level processing, building up representations of what's happening in the world around us. Participants who had previously watched the 'Sherlock' episode could anticipate events and recall them in the same order, suggesting that chu...
Researchers investigated hydro-edaphic conditions in two savanna areas, sampling 20 permanent plots and classifying 128 plant species into 34 families. The study highlights the importance of conservation in this ecosystem, which has been rapidly modified by agribusiness.
A new phylogenetic tree resolves key relationships among vertebrates, including the identification of lungfishes as closest living relatives of land vertebrates. The study uses a novel set of analyses for building large-scale genomic datasets.