Articles tagged with Data Sets
Researchers from MIT's CSAIL developed an AI system called Pic2Recipe that can analyze photos of food and predict ingredients and suggest similar recipes. The system was trained on a database of over 1 million recipes and showed impressive results, particularly with desserts like cookies or muffins.
A new dataset provides improved representation of solar variability, enabling more accurate climate model simulations and distinguishing natural from anthropogenic processes. The dataset includes enhanced estimates of solar forcing and particle effects, which may counteract rising Earth temperatures.
A new study analyzing multiple ocean datasets reveals that the oceans are robustly warming, regardless of data used. The heat redistribution among global oceans experienced a significant shift over several decades.
A new tool called Wide-Open identifies overdue scientific datasets and encourages their public release. The system uses text mining to detect dataset references in published articles and enforces open data policies.
Computer scientists and social scientists collaborate to create Collaborative Interaction Corpus (CIC) dataset, featuring 135 participants' online activity and discussion data. The dataset can be used to address various computer-science questions, including predicting user behavior and machine learning.
Researchers designed UltraTracer to work with existing algorithms, turbo-charging them for faster processing and larger datasets. The software can compare tens of thousands of neuron shapes to better understand cell types.
A new AI-based image detection method has shown promise in screening for cervical cancer, detecting abnormalities with higher sensitivity and specificity than traditional methods. The technique, developed by Lehigh University researchers, could be used to improve early detection and treatment of the disease in resource-poor regions.
Scientists at Diamond Light Source developed a new Pan-DDAA method to extract clear detail from X-ray diffraction data. The approach identifies noise sources and removes them, exploiting the ability of Diamond's beamline to repeat measurements quickly.
A recent article showcases long-term international collaboration in atmospheric sciences through annotated group photographs from two workshops, separated by over 95 years. This highlights the importance of human connection in scientific research and the value of visual datasets.
A new study published in the European Journal of Political Research reveals that women's civil rights are essential for countries to become fully democratic. The research indicates that the failure to foster women's rights compromised any attempt at democratic governance in the Arab Spring countries.
Researchers at TGen have created the largest dataset to date of extracellular small RNAs, potential biomarkers for diagnosing medical conditions like concussions. The dataset was amassed from ASU student-athletes' biofluids and helmet sensor data, aiming to develop new diagnostic and therapeutic tools.
Researchers at Umeå University have created detailed 3D maps of the insulin-producing cells in the pancreas. The data includes information on cell volume and coordinates, providing a valuable resource for studying diabetes.
A new RAND Corporation study analyzes over 200 real-world zero-day software vulnerabilities, establishing initial baseline metrics that can augment other studies. The research finds that zero-day vulnerabilities have an average life expectancy of 6.9 years, making public disclosure a moderate level of protection.
The Hyper Suprime-Cam Subaru Strategic Program has released its first public dataset, containing almost 100 million galaxies and stars. This dataset will enable scientists to explore the nature of dark matter and dark energy, as well as study the formation and evolution of galaxies.
A new study reveals that plants domesticated by pre-Columbian peoples are more likely to dominate Amazonian forests, with 85 species showing significant impacts from past human influence. The research suggests a lasting impact of human activities on plant distribution and could aid in uncovering unidentified areas of past civilization.
The HSC-SSP survey has released its first public dataset, containing 70 million galaxies and stars. The data reveals the statistical properties of dark matter, building on previous discoveries of nine clumps of dark matter in 2015.
Researchers found that resistant soybean varieties yield more than susceptible ones even at low SCN infestation levels, with a yield advantage seen in environments with no SCN infestation. Despite mounting pressure from nematodes, varieties with PI 88788-resistance still provide moderate resistance and produce good yields.
Researchers improve methods to study Earth's history by addressing bias in data from old rocks. New technique reduces errors, providing more accurate paleointensity measurements.
A team of scientists has released the largest collection of observations made with radial velocity to be used for hunting exoplanets. The dataset includes almost 61,000 measurements of more than 1,600 nearby stars and has detected over 100 potential exoplanets, including one orbiting GJ 411.
Researchers discovered a strong behavioural rule, the rule of random attraction, that explains how complex patterns of collective movement emerge in zebrafish as they develop from larvae to adults. Younger fish spend less time applying this rule, resulting in fewer schools, while adults do more, leading to group formation.
Researchers at Tel Aviv University analyzed 15,000 bat vocalizations to identify concrete evidence of socially sophisticated species that learns communication. They found that bat calls contain information about the identity of the caller and addressee, as well as specific aggressive context and possible outcome of conversation.
A statistical approach called tree bootstrapping can accurately assess uncertainty in RDS studies, enabling researchers to draw firm conclusions about vulnerable groups. This method has been applied to existing and future RDS studies, providing a basis for policy responses.
The KAT tool analyzes K-mer datasets to identify error levels, biases, and contamination in sequencing data. It also checks genome assemblies for completeness and accuracy without external reference data.
Researchers at MIT's CSAIL develop a sound-recognition system that outperforms predecessors without requiring expensive hand-annotated data. The system is trained on video and achieves high accuracy rates, with applications in improving mobile device context sensitivity and situational awareness of autonomous robots.
A new recommendation algorithm allows individuals and items to belong to multiple overlapping groups, making it more realistic than existing models. The algorithm's predicted ratings proved more accurate than those from existing systems on five large datasets.
A study published in JAMA evaluated an algorithm using deep machine learning to detect referable diabetic retinopathy, achieving high sensitivities and specificities. The algorithm was validated on two separate data sets, showing promise for increasing efficiency and coverage of screening programs.
Researchers at URV develop collaborative filtering model with scalable algorithm for predicting individual preferences and group overlaps. The new approach provides more accurate predictions than existing algorithms by considering individual differences, making it ideal for large datasets.
The need for a bridging regulatory mechanism is highlighted in e-cigarette innovations; reducing the burden on regulators while maintaining product development. A reduced data set for initial products and focused tests for variants can support this process.
A new study by Indiana University finds that bisexual individuals are paid less than their heterosexual counterparts for doing the same jobs. The research suggests that workplace discrimination may be a factor contributing to this wage gap, particularly for bisexual men and women.
Researchers at Binghamton University developed a new multilevel input layer artificial neural network to predict flight delays. The model outperformed traditional networks in terms of accuracy and training time, predicting delay lengths with about 20% more accuracy than traditional models.
Researchers at Columbia University developed a new machine-learning algorithm called TeraStructure to analyze massive genetic data sets. The algorithm can estimate population structure more accurately and twice as fast as current state-of-the-art algorithms, making it potentially useful for identifying disease-causing genetic mutations.
A UK injury data collection pilot has found a significant association between deprivation levels and an increasing incidence rate of unintentional injuries. The study suggests that collecting injury data could help reduce A&E attendances by understanding patterns in communities.
Researchers found that patients with lower severity trauma have peaks in mortality probability around 14 and 21 days after admission. The study used the largest trauma database in Europe to analyze 165,559 trauma cases, including 19,289 with unknown outcomes.
A recent study demonstrates the potential of deep reinforcement learning in optimizing traffic signal timing, reducing average delay by 14% and vehicle waiting time by 13 seconds.
Research by marketing experts found that excessive positive online customer reviews can lead to higher return rates and increased costs for retailers. This is particularly true for products with low prices or from new customers, as it raises expectations and may result in disappointment upon delivery.
Researchers have created a comprehensive dataset of non-native English sentences, providing a valuable resource for linguistic insights and practical applications. The dataset, consisting of 5,124 sentences, includes annotated errors and can help improve computers' handling of non-native English speakers.
Geographers from UC Santa Barbara's Climate Hazards Group will help African scientists predict food deficits using remote sensing tools and climate data. The new project aims to provide early warning systems for severe hunger, giving policymakers four months of advanced notice to allocate resources.
Researchers at EMBL-EBI develop an algorithm to cluster peptide mass spectra, identifying 9 million consistently unidentified spectra. This breakthrough simplifies the detection of post-translational modifications and variants, paving the way for more efficient exploitation of proteomics data.
Researchers tested facial recognition algorithms on a dataset of one million images from around the world and found that accuracy rates dropped significantly when confronted with more distractions. Google's FaceNet performed strongest, but other algorithms struggled to maintain high accuracy rates at scale.
A UCLA-led study predicts the Sierra Nevada snowpack will take years to recover from drought, even with above-average precipitation. The research provides unprecedented detail and precision, offering insights into water availability in other mountain ranges.
Researchers developed a publication strategy that allows authors to document data workflow, enabling users to reproduce simulations from scratch. This increases the reliability of simulation results and enhances model utility.
A large UK study found that married people are 14% less likely to die after a heart attack and spend two fewer days in hospital. The study suggests that physical and emotional support from a spouse can improve survival rates.
A recent 'HogWatch' survey conducted in England has provided new insights into the country's declining hedgehog population. The study suggests a moderate decline of up to 7.4% in areas frequented by hedgehogs, with possible constraints due to habitat fragmentation.
Fred Trotter, a healthcare journalist and author, has been awarded the Health Data Liberator award for his work in liberating health data. His efforts have led to the release of multiple datasets to help researchers, journalists, and companies provide data-backed healthcare solutions.
EMPIAR provides valuable data for developing new approaches to data processing and interpretation, facilitating archiving and quality control. The resource accommodates large datasets, including four related 3D SEM entries showing different stages of infection by a malaria parasite.
A new study estimates that 1.9 billion people, or 28% of the world's total population, live closer than 100 km from the coast in areas less than 100 meters above the present sea level. By 2050, this number is predicted to increase to 2.4 billion, with 500 million people living lower than 5 meters above current sea level.
Researchers used three long-term datasets to analyze honey bee population trends, identifying the USDA Census of Agriculture as the most accurate dataset for measuring changes in abundance. However, they acknowledged limitations of the data, including lack of information on colony usage and losses.
A Dartmouth College-led team has developed a new method to project future climate scenarios at the local level, addressing the limitations of global climate models in predicting climate changes at regional scales. The method generates high-resolution climate datasets for assessing impacts on small watersheds like Lake Champlain.
Researchers evaluated the performance of Riegl Z620i and LPM-800HA terrestrial laser scanners on characterizing natural surfaces. They established a procedure to test componential uncertainty budgets and demonstrated the relationship between reference network uncertainty and imaged surface repeatability.
A MIT study found that a group of consumers have an unerring knack for buying failed products, time after time. The 'harbingers of failure' tend to purchase all kinds of flop-worthy products, including those with strong initial sales but ultimately poor market reception.
A new study suggests that the beginning of the Anthropocene era was linked to the emergence of agriculture around 6,000 years ago. The researchers found a shift in the frequency of aggregated versus segregated species pairs in ancient ecosystems, which they attribute to human activity that altered the terrestrial landscape.
A study found a strong correlation between eggshell porosity and covered or exposed nest types among extant archosaurs. This suggests that eggshell porosity may be used as a proxy for nest type, allowing researchers to infer the type of nest built by extinct archosaurs.
Research found that more than half of children in the US misperceive or reject their parents' political party affiliations. Social support from parents increases the likelihood of children adopting their party identifications but not necessarily knowing them accurately.
More than half of public datasets provided with scientific papers are found to be incomplete, hindering reproducibility tests and follow-up studies. Simple errors made by researchers can compromise the process of making research data publicly available.
A Chinese research team led by Prof. Qing-Guo Huang accurately determined the Hubble constant using Baryon Acoustic Oscillation datasets, achieving a precision of around 1.3%. This result questions the measurement accuracy of Nobel laureate Adam Riess' team.
Two postdoctoral scholars from UC Santa Barbara's Kavli Institute for Theoretical Physics developed a method called ImSAnE, which constructs an atlas of two-dimensional maps for dynamic tissue surfaces. This allows scientists to analyze layered tissues with relative ease and reduces data size and processing time.
A recent study found that tourists use peer-to-peer accommodation services for social and financial reasons, leading to broader destination choices and longer stays. The availability of these services also increases tourism demand and revenue for destinations.
Genomic data sharing networks pose a risk to individual privacy if accessed by the wrong hands. Stanford researchers demonstrate a technique to prevent hacking and protect donors' identities.
Researchers used anonymous mobile phone records for 15 million people in Kenya to track the spread of rubella, predicting seasonal disease patterns. The study shows that cellphone movement can be a predictor of infectious-disease spread, particularly among children during school breaks.
A new genomic catalog of nearly 200 viral genomes provides insight into the origins and transmission of Lassa virus. The data set confirms rodent-to-human transmission as the primary mode of infection.