Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
The American Society for Biochemistry and Molecular Biology annual meeting will explore cutting-edge approaches in biochemistry and molecular biology. Leading experts will discuss the latest discoveries and advancements in fields such as AI, RNA regulation, and carbohydrates in health and disease.
Researchers at Columbia University are developing algorithms that enable robots to understand object permanence and learn from 3D information. This allows robots to track objects and humans as they move around, improving their perception capabilities in indoor environments.
Machine learning models can automatically detect individual heads on grain crops in images taken from drones, providing a simpler alternative to manual counting. The study provides a detailed pipeline outlining the use of these models, covering data preparation, model validation, inference, and yield-specific metrics.
A new study led by Yale School of the Environment researchers examines the application of machine learning in the sustainable development of biomass. The study found that machine learning has not been applied across the entire life cycle of biomass-derived materials (BDM), limiting its ability to advance research and development.
Researchers used AI to map sparse life hidden in salt domes, rocks, and crystals at Salar de Pajonales, a Martian analog. The study found that microbial life is concentrated in patchy biological hotspots linked to water availability, and AI can detect biosignatures up to 87.5% of the time.
A new machine learning model developed by NYU researchers can predict food crises up to 12 months in advance by analyzing news articles and their frequency. The model shows a high correlation between news coverage and on-the-ground occurrences of risk factors, indicating its potential as an early-warning system.
A POSTECH research team developed a machine learning model that accurately predicts cancer type-specific driver mutations, shedding light on distinct pathological mechanisms across various tumors. The model's performance outperformed current leading methods of detection, with better accuracy and sensitivity.
Researchers at Massachusetts General Hospital developed an accurate method for detecting Alzheimer's disease using routinely collected clinical brain images. The AI model detected Alzheimer's disease risk with 90.2% accuracy across five datasets, regardless of age or scanner manufacturer.
GPT-3 performs nearly on par with humans in decision-making but struggles with causal reasoning and information search. The language model's limitations may be due to its passive information-gathering approach, highlighting the need for active interaction with the world to achieve human-like intelligence.
A new study published in PLOS One demonstrates the potential of machine learning and artificial intelligence to aid in the development of new treatments for Rett syndrome. The researchers used a wearable electronic chest patch to monitor cardiac activity and movement, and developed an algorithm that identified patterns specific to seve...
A new machine learning model can predict city traffic activity in different zones of cities, enabling targeted responses from policymakers. Understanding people's mobility patterns is crucial for improving urban traffic flow, and the model provides insights into urban interactions.
Bielefeld University researchers developed an AI method using Capsule Networks to analyze genotype profiles of 3,000 ALS patients, achieving 87% accuracy in predicting whether or not people will develop ALS. The study reveals over 900 genes that play a role in identifying the disease.
A team of researchers from Rensselaer Polytechnic Institute has developed a system to optimize TV ad scheduling, resulting in a 3-5% revenue increase for networks. The model combines mathematical programming and machine learning to assign ads to specific breaks and positions.
Researchers create lab-grown brain organoids as biological hardware to drive computing forward. These three-dimensional cultures can process information faster and more efficiently than current silicon computers.
Researchers at Sandia National Laboratories and Purdue University have developed a technique called moving target defense to protect against hackers taking control of military jets. This approach involves shuffling network addresses to make it difficult for attackers to access critical systems.
The K-EMERGE project aims to capture data from textbooks and human experts, combining them using deep-learning AI-based Natural Language Processing (NLP) and cognitive architecture modeling. The system learns continuously to evaluate and enhance the integrity of the data.
Researchers found that individuals who scored high on questionnaires assessing mystical and insightful nature of their psychedelic experiences reported improvements in anxiety and depression symptoms. A challenging experience while on these substances was also beneficial, especially in the context of mystical and insightful experiences.
A novel ensemble learning model has achieved an accuracy of 96 percent in predicting mild cognitive impairment and dementia in older drivers. The two most influential driving variables are the right to left turn ratio and the number of hard braking events.
A new cybersecurity framework uses digital twin technology, machine learning, and human expertise to detect cyberattacks in manufacturing processes. The framework analyzes continuous data streams from physical machines and their digital twins to identify irregularities and flag potential threats.
Researchers used machine-learning algorithms to design new light-emitting enzymes called luciferases that can efficiently recognize specific chemicals and emit light. This breakthrough could lead to custom enzymes for a wide range of applications in biotechnology, medicine, environmental remediation, and manufacturing.
The University of Minnesota Medical School has received $1.4 million in funding to develop methods and risk management processes for clinical artificial intelligence and machine learning to optimize patient safety. The team will focus on understanding the performance of clinical AI at the individual patient level.
Scientists at IISc develop neuromorphic camera that uses machine learning to pinpoint objects smaller than 50 nanometers in size, enabling nanoscale precision in biological processes, chemistry, and physics. The technique combines optical microscopy with the neuromorphic camera and machine learning algorithms.
Computer scientists designed a reconstruction attack that proves US Census data can be exposed and stolen with current privacy measures. The study demonstrates risks to individual respondents' privacy, highlighting the need for differential privacy techniques to protect sensitive information.
A new study by Tulane University demonstrates that even a single atom can act as a reservoir for computing, processing information optically. The researchers proposed a non-linear single-atom computer where input and output are encoded in light, enabling flexible computation with any desired outcome.
Researchers developed a machine learning model using high-resolution satellite imagery to estimate aboveground carbon stocks in the Amazon. The study found that accounting for uncertainties in forest degradation classification led to lower estimates of mean carbon density, suggesting earlier estimates may have been over-optimistic.
A Penn State-led research team found that language models can plagiarize content in three ways: verbatim, paraphrase, and idea reuse. The study highlights the need for more research into text generators and their potential ethical implications.
Researchers have developed a hand-held device that can rapidly and accurately identify lesions in the mouth that will develop into cancer. The Liverpool Diagnostic Infrared (LDIR) Wand employs infrared lasers to predict future cancer risk based on machine learning analysis.
Researchers used machine learning to classify hundreds of thousands of X-ray objects, discovering thousands of new cosmic objects including black holes and neutron stars. This breakthrough establishes a state-of-the-art capacity for applying machine learning techniques in fundamental astronomy research.
A new technique maps the effects of fire-induced permafrost thaw in Alaska, revealing widespread topographic change and vegetation shifts. The study used a machine learning-based approach to quantify thaw settlement across 3 million acres of land, with results showing a significant loss of evergreen forest and shrubland encroachment.
A new machine learning model combines fusion gene profiling, serum PSA level, and Gleason score to predict prostate cancer recurrence with improved accuracy. The model outperformed clinical data alone and provided valuable insights into the mechanism of disease progression.
Researchers at MIT developed a technique to improve machine-learning models' reliability without requiring additional data or extensive computing resources. The method uses a simpler companion model to estimate uncertainty, enabling more effective uncertainty quantification.
Researchers at the University of Gothenburg have developed three AI-based decision support systems for cardiac arrest care, which can help doctors identify key factors affecting patient outcomes. The tools are based on large datasets and provide accuracy rates of up to 95% in predicting patient survival or death.
A new study uses Fourier analysis to understand how deep neural networks learn complex physics. By analyzing the equation of a fully trained model, researchers were able to identify crucial information about how the network learns and generalizes. This breakthrough could accelerate the use of scientific deep learning in climate science.
Rice University researchers have developed an innovative system to study mosquito feeding behavior using fake skin made with a 3D printer, eliminating the need for live volunteers. The system was tested on various mosquito repellents and showed promising results, suggesting it could be scaled up for future studies.
Researchers have developed a new synthetic skin, made of hydrogels, to study how mosquitoes transmit deadly diseases. The hydrogel system can mimic different blood vessel patterns, allowing for more consistent testing and analysis. This breakthrough may help identify ways to prevent the spread of disease.
Researchers applied machine learning tools to study how climate impacts connectivity and biodiversity in the Pacific Ocean's Coral Triangle. They found that climate dynamics have contributed to biodiversity due to variability introduced by El Niño and La Niña events.
Researchers used Reinforcement Learning to enable kites and gliders to adjust their orientations in real-time, accounting for turbulence. This improvement could significantly enhance the performance of airborne wind energy devices, expanding the reach of wind power to poorer communities.
FRIDA uses AI models similar to those powering tools like ChatGPT and DALL-E 2 to generate paintings based on user input. The robot's final products are impressionistic and whimsical, with bold brushstrokes that lack precision sought in robotic endeavors.
Researchers utilized the Chemistry42 platform to generate novel molecular structures and identified a hit molecule for CDK20, a promising target for hepatocellular carcinoma. The platform's customizable reward function and generative models enabled efficient design and optimization of molecules.
New research from the University of Georgia reveals that artificial intelligence can be used to find planets outside our solar system. Machine learning can analyze environments where planets are still forming, helping scientists overcome difficulties such as distance and data thickness.
The iCPH platform combines physical and cyber elements to capture human motions, using musculoskeletal analysis and machine learning. It generates contact motion networks for humanoid robots and simulates human behaviors, enabling smooth interactions with humans.
Researchers developed a machine learning model using advanced 2D chemical descriptors to predict highly selective asymmetric catalysts without quantum chemical computations. The model demonstrated high accuracy in predicting catalyst structures and selectivity, outperforming existing methods.
Researchers have developed a sophisticated AI algorithm, SPHINKS, that can refine omics datasets and pinpoint protein kinases responsible for tumor growth in glioblastoma. The algorithm has the potential to provide personalized treatments for patients with aggressive brain cancer.
Researchers developed BirdFlow, a predictive model that utilizes eBird data and machine learning to forecast migratory patterns. The model was tested on 11 species of North American birds and found to outperform other models in tracking migration flows.
A new approach to deep reinforcement learning demonstrates ability to stabilize large datasets used in AI models, which may lead to uncovering ways to arrest cancer development. The method has been successful in designing and refining existing therapies, with the next step being to use live cells.
Researchers have developed an AI-based resource to assist individuals in identifying recommended actions based on their clinical profile and COVID at-home test results. The system uses a combination of symptoms and home tests to provide more accurate diagnoses and improve patient care.
Researchers at Bar-Ilan University have developed a new type of artificial neural network that outperforms traditional deep learning architectures. By using tree architecture with single routes to output units, they achieve better classification success rates, paving the way for efficient and biologically-inspired AI hardware.
Dr. Nico Spiller to develop new analysis methods using machine learning to analyze complex brain data related to memory, decision making, and movement. The fellowship aims to provide insights into neurodegenerative diseases such as Alzheimer's and Parkinson's disease.
NeuralTree is a closed-loop neuromodulation system-on-chip that can detect and classify biomarkers from real patient data and animal models of disease in-vivo, leading to high accuracy in symptom prediction. The system boasts 256 input channels, making it highly versatile and scalable.
Researchers applied deep learning techniques to a previously studied dataset of nearby stars, uncovering eight previously unidentified signals of interest. The new approach enabled faster and more accurate results, with the potential to accelerate discovery of extraterrestrial life.
Scientists developed an AI system, ProGen, that can generate artificial enzymes from scratch, working as well as those found in nature. The AI model learned aspects of evolution and was able to tune its generation for specific effects, creating proteins with unique properties.
Researchers created a system to monitor underground gas pipelines using high-tech sensors that can detect weaknesses, discrepancies, and diversion in residential natural gas lines. The method uses ultrasonic sensors to transmit signals through the pipe, limiting the likelihood of gas diversions and ensuring public safety.
A new project aims to examine the circulation of newspaper reports on anti-Black violence between 1863 and 1921. The team will use computational methods to trace how stories spread across the country and map their impact, with potential applications for studying other forms of racial violence.
A $2.3 million grant from the US Department of Energy funds a 'solar testbed' at I-79 Technology Park in Fairmont, supporting research on battery storage, grid integration, and cybersecurity. The project aims to assess solar panel health and monitor grid interactions with solar power.
Researchers modified an algorithm to detect urinary tract infections (UTIs) in primary care settings, removing microscopy features that weren't available. The new algorithm performed well and suggests withholding antibiotics from low-risk patients to reduce antibiotic overuse.
Researchers successfully applied AlphaFold AI to an end-to-end platform, discovering a novel target and developing a potent hit molecule for liver cancer. The study demonstrates the potential of AI-powered drug discovery to accelerate treatment development.
Researchers at the University of Wisconsin-Madison have developed a machine-learning model that detects cancers at an early stage by analyzing fragments of cell-free DNA in plasma. The technique, which uses readily available lab materials, distinguished people with any stage of cancer from healthy individuals 91% of the time.
Researchers have developed a machine learning model that can predict the word about to be uttered by a subject based on their neural activity. The model achieved 55% accuracy using six channels of data and 70% accuracy using eight channels, comparable to other studies requiring electrodes over the entire cortical surface.
The study, published in Nature Medicine, demonstrates the first-ever use of federated learning to train deep learning models on histopathology data from multiple hospitals without compromising data privacy. This breakthrough has the potential to unlock precision medicine through secure and AI-powered medical research.
MIRMI researchers create robotic waiter with precise control using the principles of a spherical pendulum, achieving 'slosh-free movement' and improving safety. The solution has potential applications in healthcare and hazardous materials transport.