Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
A new study uses machine learning to search for antibiotics in a vast dataset of microbial genomes, identifying 863,498 candidate antimicrobial peptides. Promising results are observed in initial tests against disease-causing bacteria and preclinical animal models.
A new study by the Society for Risk Analysis explores the impact of AI-driven cyberattacks on global economies, supply chains, and trade. The research found that these attacks can cause significant declines in real GDP, trade prices, and volumes, as well as disruptions to trade routes, particularly among heavily reliant digital economies.
Researchers found human infants use 'helpless' period to pre-train brain, leading to rapid learning and high performance, similar to machine learning models. This study challenges classic explanation for infant helplessness and could inspire next gen AI models.
A team of researchers from Princeton University and the US Department of Energy's PPPL have successfully deployed machine learning methods to suppress harmful edge instabilities in fusion devices. Their approach optimizes the system's suppression response in real-time, maintaining high plasma performance without sacrificing stability.
A new method for detecting defects in additively manufactured components uses deep machine learning, generating synthetic defects for training and testing on physical parts. The algorithm accurately identifies hundreds of defects, even those unseen by the model before.
The team created a prediction model that generates sustainable products with high accuracy and explores the vast design space of aerogel assembly. Their strong and flexible aerogels have programmable mechanical and electrical properties, opening up new possibilities for green technologies.
A comprehensive study led by Dr. Luan Shenghua of the Chinese Academy of Sciences found a general factor of impulsivity that is stable and predictive of behaviors, contradicting claims of its demise as a personality trait.
A new open-source platform called CheckMate allows users to interact with and evaluate the performance of large language models (LLMs) like ChatGPT. Researchers found that while LLMs can be helpful, they also make mistakes and provide incorrect information.
Researchers developed a machine learning model that predicts ideal oxygen levels for individual patients based on characteristics such as age, sex, and heart rate. The results suggest personalized oxygenation targets could reduce mortality rates, offering new hope for critical care patients.
Portland State University has secured a nearly $1 million grant from the National Science Foundation's Campus Cyberinfrastructure program to establish the Oregon Regional Computing Accelerator (Orca) cluster. The cluster will provide free-of-cost computing resources and cyberinfrastructure to colleges in rural, regional, and minority-s...
Researchers at La Jolla Institute for Immunology developed a computational method to link gene activity to molecular marks on DNA, potentially aiding in the detection of solid tumors and more accurate cancer diagnoses. This new approach utilizes machine learning tools to identify connections between genes and enhancers in the genome.
Researchers at Texas A&M University are investigating the historical effects of strain on shape-memory alloys to improve predictive capabilities. They will use a synergistic experimental and numerical approach to understand and predict history effects in these alloys, with potential applications in heart stents and airplane wing flaps.
A novel approach to training AI systems uses information about spatial position to identify objects and navigate surroundings, inspired by children's visual development. The method improves contrastive learning models' effectiveness by incorporating simulated spatial context information, outperforming base models in various tasks.
A new study from Chalmers University of Technology shows that AI-controlled charging stations can offer personalized prices to electric vehicle users, minimizing both price and waiting time. However, the researchers highlight the importance of addressing ethical issues related to data exploitation by motorists.
Researchers Dr. Samson Zhou and Dr. David P. Woodruff aim to create secure algorithms for big data models using mathematical connections and cryptography ideas. They focus on streaming models, which process data in real-time, and address challenges such as randomness and different types of attacks.
A team of researchers from Rice University and the University of Michigan found that some neurons not only replay recent past experiences but also anticipate future experience during sleep. The discovery provides an unprecedented view of how individual neurons in the hippocampus stabilize and tune spatial representations during periods...
Researchers developed a new method to enhance thermal image super-resolution by employing synthetic imagery, significantly improving detail and utility of thermal imaging across various applications. The approach utilizes high-resolution images from the visible spectrum to guide the super-resolution of low-resolution thermal images.
Researchers at Duke University developed an assistive machine learning model that greatly improves the ability of medical professionals to read EEG charts. The model, which provides visual explanations and decision support, has been shown to almost double medical professionals' accuracy in identifying seizure-like events, potentially s...
A team of researchers from Japan, China, and Finland created a system called generative content replacement (GCR) that uses AI to replace parts of images that might threaten confidentiality with visually similar but AI-generated alternatives. In tests, 60% of viewers couldn't tell which images had been altered.
Researchers have developed a system combining bio-inspired cameras with AI to quickly detect obstacles around cars, using less computational power. The hybrid system detects objects up to one hundred times faster than current systems while reducing data transmission and processing needs.
Researchers have developed a method to detect microplastics in marine and freshwater environments using porous metal substrates and machine learning. The system can identify six types of microplastics with high accuracy, offering a cost-effective solution for environmental monitoring.
A new deep learning AI model, Dev-ResNet, identifies embryonic developmental events in pond snails using video analysis. This breakthrough enables the detection of key features, such as heart function and hatching, with unprecedented sensitivity.
Researchers at Oregon Health & Science University have received a $16.4 million grant to advance mental health care for children, leveraging machine learning and novel clinical measures to improve prediction, diagnosis, and treatment of mental health conditions across childhood and adolescence. The project aims to create an actionable ...
The team aims to create a system that can deliver items without human contact, using cables, knots, and multiple robots. They will focus on scaling up the transport of small objects like a basketball and solar panel.
Researchers have developed an AI algorithm that can track protein clumping under the microscope in real-time, revolutionizing the study of neurodegenerative disorders. The tool helps identify key characteristics of clumped proteins, which can lead to new therapies.
Scientists from TU Delft and Brown University engineer string-like resonators capable of vibrating for extended periods at room temperature, enabling sensitive sensing applications. The innovation uses advanced nanotechnology techniques and machine learning algorithms to create ultra-long strings with minimal energy loss.
Researchers have developed a machine learning model to identify high-performance multicomponent metal oxide electrocatalysts for the oxygen reduction reaction. The study found that certain features, such as itinerant electrons and configuration entropy, are critical for achieving high current density in ORR.
A new AI model developed by Cold Spring Harbor Laboratory's Benjamin Cowley and team uses a 'population code' to predict fruit fly behavior, revealing that multiple neurons combine to sculpt actions. The breakthrough enables the AI to accurately predict how real flies will behave in response to visual stimuli.
A US Army research collaboration with Boston University's KABlab used an AI machine learning robot to create a record-breaking energy-absorbing shape, breaking the known record of 71% efficiency. The shape has four points, like thin flower petals, and is taller and narrower than early designs.
Researchers at the University of Innsbruck developed a novel method using diffusion models to generate quantum circuits. The model can produce accurate and flexible circuits, including those tailored to specific quantum hardware connections.
Engineers developed a material that mimics human bone for orthopedic femur restoration, providing optimized support and protection from external forces. This innovative approach uses machine learning, optimization, and 3D printing to create a fully controllable computational framework.
Researchers applied machine learning models to predict groundwater depth in Ningxia, China, achieving better performance than traditional methods. The hybrid models outperformed multiple linear regression, and the DBO algorithm further enhanced prediction accuracy.
The DGIST-Stanford joint research team successfully developed a novel medical AI model based on federated learning, which can accurately segment body organs by effectively learning medical image data from different hospitals. The technique uses shared embedding learning to enable federated learning without data breaches and leaks.
Researchers at Nagoya University found that cooperative hunting does not require complex cognitive processes, but rather simple rules and experience. The study used computational models and simulations to demonstrate the effectiveness of cooperation in hunting, with AI agents learning to work together through reinforcement learning.
Model disgorgement is a set of techniques that force generative models to remove content leading to copyright infringement or biased responses. Researchers propose this approach to address issues like stylistic infringement, where models reproduce copyrighted works in the style of famous artists.
Researchers from MIT and the University of Basel developed a physics-informed machine-learning framework that can automatically map out phase diagrams for novel physical systems. This approach leverages generative models, making it possible to detect phase transitions without requiring huge training datasets. The technique has potentia...
A new approach uses artificial intelligence to turn low-quality images into high-quality ones, enhancing the image quality of metalens cameras. This technology could make these cameras viable for intricate microscopy applications and mobile devices.
Researchers used D-FFOCT and deep learning to create an intraoperative diagnostic workflow for breast cancer patients. The approach achieved high accuracy and speed, reducing processing time by a factor of 10 compared to conventional histology.
PPPL researchers utilize machine learning to perfect plasma vessel design, optimize heating methods, and maintain stable control of fusion reactions. The team achieves significant results by predicting disruptions and adjusting settings before instabilities occur, enabling high-confinement modes in tokamaks.
Researchers developed an approach combining quantum mechanical density functional theory and artificial intelligence to predict high-temperature superconducting materials. Over 120 structures with superior properties were found, including comparison to MgB2 at 39 K.
By recasting diffusion as a sum of individual contributions called 'kinosons,' researchers developed a new method to model alloy behavior. Machine learning is used to compute the statistical distribution of these contributions, allowing for fast and accurate simulation of diffusion. This breakthrough enables significant improvements in...
Insilico Medicine's lead compound demonstrates strong enzymatic activity, selectivity, and favorable ADME properties, as well as antitumor activity in various animal models. The company's generative AI-powered platform generated over 3,600 candidate molecules before identifying the promising lead compound.
Researchers at Concordia University developed a novel framework to detect counterfeit coins by analyzing image features and patterns. The method uses fuzzy association rules mining and can be applied to detect other types of counterfeit items, such as fake goods and labels.
Researchers at Yokohama National University utilized machine learning and AI to predict the selectivity of chemical reactions. By analyzing molecular factors such as sterics and orbitals, they developed a method to better understand reaction mechanisms, leading to more efficient synthesis of desired products.
A high school student, Michelle Du, helped develop a novel method to predict neurotransmitters from insect connectomes using neural networks. The method has been used in various neuroscience studies and provides valuable insights into brain circuit function.
Researchers created a digital twin model that predicts and controls complex systems, achieving higher accuracy than traditional methods. The algorithm is compact, energy-efficient, and easy to implement, making it suitable for self-driving vehicles and other dynamic systems.
Researchers have developed a new imaging technique that rapidly and accurately identifies cancerous tissues in breast samples. The method uses machine learning algorithms trained on hyperspectral dark-field microscopy data to pinpoint regions of invasive ductal carcinoma and invasive mucinous carcinoma.
Researchers developed ChemCrow, an AI-powered tool that integrates expertly designed software tools to autonomously perform chemical synthesis tasks. The system enables plan-and-execute approach with reduced hallucinations and practical application, accelerating research and development in pharmaceuticals and materials science.
A team at the University of Münster developed an improved method for explaining machine predictions of chemical reactions, using mechanisms such as reproduction, mutation and selection. The algorithm creates customised molecular fingerprints that predict chemical reactions with surprising accuracy, suitable for predicting quantum chemi...
A study by Albert-László Barabási and colleagues used machine learning to identify phrases representing allusions to foundational papers in the physics literature. The findings suggest that hidden citations obscure true impact in science, with influential ideas becoming so familiar that researchers stop citing their sources.
Researchers created GraSSRep and rhea, tools that outperform current methods for handling repeats and structural variants in metagenomic data. These methods use self-supervised learning and graph neural networks to analyze microbiome data, offering new insights into biological processes and potential applications in antibiotic resistance.
Researchers from GIST and MIT CSAIL develop a MultiSenseBadmpton dataset to enhance badminton performance analysis and AI-based coaching. The dataset captures joint movements, muscle signals, and gaze movements of badminton players.
Researchers at Iowa State University are developing an AI tool that can identify agricultural pests from photos, including insects and weeds. The project will use a supercomputer to train ensemble models that can analyze images quickly and provide pest-control strategies.
A new AI model developed by researchers at the University of British Columbia can accurately predict if a patient receiving cancer care will require mental health services. The AI analyzes oncologist's notes and identifies subtle clues that suggest a patient may benefit from early psychiatric or counselling interventions.
Researchers found a strong association between favorable survival outcomes and high populations of tissue-resident memory T cells in melanoma patients. The study identified 11 distinct gene signatures that correlate with T cell abundance and patient survival, suggesting a crucial role for T cells in immunomodulation.
Researchers developed a new AI algorithm called Maximum Diffusion Reinforcement Learning (MaxDiff RL) to improve robot reliability. The algorithm enables robots to learn complex skills more efficiently by encouraging exploration of their environments.
A massive open dataset, OpenDAC, has been created to accelerate direct air capture technology development while reducing costs. The database enables the training of an AI model that predicts material interactions with high accuracy, significantly faster than traditional chemistry simulations.
A new AI method developed by Swedish researchers can identify toxic substances based on their chemical structure, potentially replacing animal testing. The method has been shown to be more accurate and broadly applicable than existing computational tools, offering a promising alternative for environmental research and authorities.
FathomVerse combines real underwater images, interactive gameplay, and AI training to improve ocean research. The game engages players in labeling images for machine learning models, contributing to a broader audience in ocean exploration.
Researchers designed a new AI model that accurately predicts treatment outcomes for stroke prevention in people with heart disease. The model, called CURE, was trained on a massive dataset of de-identified patient information and fine-tuned to focus on specific health conditions and treatments.