Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
Researchers at Virginia Tech have designed a new metallic material alloy with superior mechanical properties, leveraging data-driven frameworks and explainable AI. This breakthrough accelerates the discovery of advanced metallic alloys, offering insights into materials' structure-property relationships.
Researchers at Duke University have developed a new framework called HUMAC that enables robots to collaborate like humans by teaching them Theory of Mind. After just 40 minutes of guidance, robot teams exhibited strong collaborative behaviors and achieved high success rates in simulations and physical tests.
Developers of educational tools focus on technical challenges, while educators are concerned with broader impacts, such as inhibiting critical thinking skills and exacerbating systemic inequality. Researchers recommend designer-centered approach to facilitate development of educator-centered edtech.
Researchers found that vision-language models fail to understand negation words like 'no' and 'not', which can cause incorrect diagnoses. This limitation affects multiple choice question answering with negated captions, with accuracy dropping below random chance.
New York University researchers developed a novel process using machine learning to reveal groups of genes governing nitrogen use efficiency in plants like corn. The study aims to help farmers improve crop yields and minimize fertilizer costs.
A new AI-driven tool can forecast acute child malnutrition in Kenya up to six months in advance with high accuracy, enabling timely interventions. The model integrates clinical data and satellite information to identify emerging risk areas, providing a game-changing solution to address public health emergency in the country.
A University at Buffalo-led study proposes using AI-powered handwriting analysis to identify spelling issues, poor letter formation, and other indicators of dyslexia and dysgraphia. The work aims to augment current screening tools and provide an early detection tool for these neurodevelopmental disorders.
A new deep learning model, ENDNet, significantly enhances subgraph matching accuracy by identifying and neutralizing extra nodes that interfere with the matching process. This improves performance in pattern recognition tasks across various fields, including drug discovery and natural language processing.
A new AI model can predict protein location in human cells with high accuracy, enabling faster diagnosis of diseases like Alzheimer's and cancer. By combining protein sequence analysis with computer vision, the model can pinpoint proteins' locations at the single-cell level.
A new study uses explainable machine learning to predict metastasis and mortality in children with medulloblastoma, identifying key immune factors driving risk. The model provides a transparent approach to prognostic accuracy and supports personalized clinical decision-making.
A study found that large language models generalize language patterns in a surprisingly human-like way through analogy, relying on stored examples and drawing analogies when dealing with unfamiliar words. The models behave as if they have formed a memory trace from every individual example of every word encountered during training.
A new AI-based method accurately sorts cancer patients into groups with similar characteristics before treatment and similar outcomes after treatment. The approach showed better performance than standard statistical and machine learning methods in predicting treatment outcomes from health record data.
Researchers developed an AI-based music creation support system called Amuse, which converts user inputs into harmonic structures to support composition. The system has high potential as a creative companion for musicians and is centered on the creator's initiative.
A team of researchers at Rice University developed a new strategy for identifying hazardous pollutants in soil using light-based imaging and machine learning algorithms. The approach can detect toxic compounds like PAHs and PACs even when no experimental data is available, addressing a critical gap in environmental monitoring.
A Carnegie Mellon University researcher critiques an article on large language models' ability to analyze sensitive discourse, including hate speech. The commentary highlights the need for a nuanced approach to human-AI collaboration in critical discourse analysis.
Researchers employed Bayesian neural networks to fit photonuclear cross-sections with remarkable reliability, outperforming traditional methods like TENDL-2021. The approach demonstrated superior accuracy in describing low-energy thresholds and high-energy tails, particularly for sparse or biased data.
A new study from the University of Surrey explores how metaverse platforms like Roblox and ZEPETO are changing consumer engagement forever. Digital doppelgängers, using AR and VR technologies, create immersive experiences that enhance brand engagement and emotional connections.
A new microscopy method, LICONN, developed by ISTA scientists and Google Research, can reconstruct mammalian brain tissue with all synaptic connections between neurons. This technique uses standard light microscopes and hydrogel to achieve high resolution and opens up possibilities for visualizing complex molecular machinery.
Researchers have discovered nearly 35,000 publicly downloadable deepfake image generators, downloaded over 15 million times since late 2022, primarily targeting women. The study highlights the need for robust technical safeguards and regulatory approaches to address the creation and distribution of these harmful AI models.
Large Language Models (LLMs) like Mixtral can grade student responses quickly, but often rely on shortcuts and assume students understand topics. Researchers found that LLMs are more accurate when provided with human-made rubrics, which include specific rules for grading. This suggests that AI can be used to streamline grading processe...
A new study from McGill University suggests AI can revolutionize conservation efforts by enabling data-driven decision making. The research identifies key ways AI can close biodiversity knowledge gaps, including species identification and distribution mapping.
Researchers have used machine learning to study the melting of layered materials, discovering a complex two-step process that contradicts prior theories. The team identified changes in topological excitations as the key to understanding the unexpected melting behavior, enabling predictions up to 12 material layers.
The journal is inviting submissions on AI applications in biomedical engineering, focusing on medical device development, personalized diagnostics, and patient outcome prediction. Accepted articles will be published as part of a themed section showcasing current research on AI applications in biomedical engineering.
A new study published in the Journal of the National Cancer Institute suggests that artificial intelligence can help detect interval breast cancers earlier, potentially reducing their rates by 30%. The research used AI software to analyze mammograms and identify subtle signs of cancer that were missed by radiologists.
Researchers used generative AI to design diverse mitochondrial targeting sequences, achieving a 50-100% success rate in yeast, plant cells, and mammalian cells. The AI-generated sequences showed improved targeting abilities compared to existing ones, with potential applications in metabolic engineering and therapeutics.
A recent study by the University of Surrey suggests that hotel managers may need to adapt to AI-driven system management, shifting their focus from controlling to coaching staff. Effective communication, emotional intelligence, and creativity are key strategies for managers to navigate this transformation.
Researchers at MIT developed a technique to improve the reliability of conformal classification, which can produce impractably large prediction sets. By combining test-time augmentation with conformal prediction, they reduced prediction set sizes by up to 30 percent while maintaining probability guarantees.
A Dartmouth team's AI model recognized Navajo with near-perfect accuracy, identifying related languages such as Apache and Native Alaskan languages. The study suggests that this technology could be a bridge to including smaller languages in online translation services.
Researchers developed an AI system to classify tiny powdery pollen grains produced by fir, spruce, and pine trees, enhancing speed and accuracy. The tool can aid allergy sufferers, urban planners, farmers, and wildlife conservation efforts.
Researchers developed a novel vote-based model for accurate hand-held object pose estimation, addressing issues with existing approaches. The new framework achieves significant improvements in accuracy and robustness, enabling robots to handle complex objects and advancing AR technologies.
A team of researchers used machine learning to analyze changes in astrocyte cells' structure, shedding light on heroin addiction and relapse. The study, published in Science Advances, found that specific subpopulations of astroglia exhibit more pronounced morphological changes during drug use.
A new study finds that AI tools can significantly improve the readability of online patient education materials, making them more accessible for patients. The researchers used three large language models to optimize the readability of materials without compromising accuracy, resulting in improved scores and reduced word counts.
A study by University of East Anglia compared 145 real student essays with 145 ChatGPT-generated ones, finding that AI essays were coherent but lacked engagement markers like questions and personal commentary. This reflects the limitations of AI in replicating human writing's conversational nuance.
A new review advocates for building confidence in AI applications by implementing robust data governance frameworks, enhancing transparency, and involving stakeholders. The authors emphasize the importance of addressing ethical implications and ensuring equitable access to AI-driven innovations in clinical oncology.
Kavraki's interdisciplinary research in robotics and biomedicine has been recognized for its impact on manufacturing, space exploration, and medicine. Her work bridges theory and application, with contributions to novel robot motion planning, personalized cancer treatments, and drug discovery.
César A. Uribe, a Rice University professor, has received an NSF CAREER Award to develop mathematical tools for decentralized learning in AI and data science. His research aims to create more efficient networks of computers that can process massive amounts of data without relying on centralized coordination.
Researchers have developed a neuro-quantum leap in finding optimal solutions, leveraging Fowler-Nordheim annealers to discover new and unknown solutions. NeuroSA's structure is neuromorphic, with a search behavior determined by FN annealer, making it powerful for solving complex optimization problems.
Researchers develop mathematical modeling to predict aflatoxin outbreaks in Texas using remote sensing satellites and soil properties. The model has the potential to save farmers billions of dollars in losses by providing early risk prediction and targeted prevention strategies.
Researchers at Rice University developed a new machine learning algorithm that excels in interpreting light signatures of molecules, materials and disease biomarkers. The tool can detect subtle signals in optical spectroscopy, enabling faster medical diagnoses and sample analysis.
Researchers found that pretrained foundation models outperformed a standard baseline model in identifying nonmelanoma skin cancer from digital images. The models achieved accuracies of up to 92.5% in distinguishing between NMSC and normal tissue, significantly improving upon an older architecture.
A new study found that artificial intelligence can better estimate the prevalence of physical abuse in children seen in emergency departments. The AI model correctly identified child abuse in over 35% of cases, outperforming traditional methods that relied solely on diagnostic codes entered by providers or administrative staff.
Researchers identified three subtypes of senescent skin cells with distinct shapes, biomarkers, and functions. The study found that one subtype, C10, becomes more prevalent with age and is harder to kill with existing drugs.
Researchers at ETH Zurich developed a method to specifically reduce uncertainty in AI responses by enriching general language models with additional data from relevant subject areas. The SIFT algorithm uses relationship vectors to identify closely related information, resulting in more reliable answers and reduced computational overhead.
The Simons Collaboration on Ecological Neuroscience (SCENE) is a 10-year program that will support projects aiming to uncover how the world shapes representations in the mind and brain. By integrating sensory and motor information, researchers hope to discover fundamental principles of cognition applicable across species.
Researchers developed an AI model called Odor Generative Diffusion (OGDiffusion) to automate fragrance creation, generating essential oil blends based on user input of scent descriptors. The system produced fragrances that met people's expectations in human sensory tests.
Machine learning (ML) techniques can identify materials with high synthesis feasibility and suggest suitable experimental conditions. Computational models derived from thermodynamics and kinetics enhance predictive performance and interpretability of ML models, optimizing experimental design and increasing synthesis efficiency.
A new study has found that combining signals from electromyography and force myography can improve the accuracy of prosthetic control. The researchers used a combination of these two techniques to classify hand gestures with high accuracy, outperforming both methods alone.
MIT researchers have created a unifying framework that combines existing ideas to improve AI models or create new ones. The 'periodic table of machine learning' categorizes classical algorithms based on the approximate relationships they learn, allowing for fusion of strategies and discovery of new algorithms.
A new robotic framework allows robots to learn tasks by watching a single how-to video, significantly reducing the time and energy needed for training. The RHyME system enables robots to adapt to real-world environments and perform multiple-step sequences with improved success rates.
Mass General Brigham researchers developed a machine learning algorithm, PAMmla, to predict properties of genome editing enzymes. The approach helps reduce off-target effects and improves editing safety and efficiency, enabling customized enzymes for new therapeutic targets.
Researchers developed a new method that uses simple grayscale eye photos to predict anemia in children. The technique analyzes patterns and textures in the conjunctiva of the eye, avoiding problems caused by different light conditions or camera models.
A new machine learning model accurately predicts the fitness of AAV capsids based on their amino acid sequence, enabling more efficient and cost-effective gene therapies. The model's robustness and generalizability have been demonstrated through tests on independent datasets, offering a promising tool for capsid engineering.
Researchers developed an intelligent lithium plating detection system using a Random Forest machine learning algorithm, analyzing pulse charging data to identify subtle electrical signatures. The system achieves high accuracy and can be implemented without modifying existing battery systems.
A new smart insole system monitors how people walk in real time to improve posture and provide early warnings for conditions like plantar fasciitis and Parkinson’s disease. The system offers high-resolution spatial sensing, self-powering capability, and combines with machine learning algorithms.
A team of researchers at Kyoto University has developed a simple but effective method for detecting early wood coating deterioration, which can extend the life of wooden structures and improve sustainability. The approach combines mid-infrared spectroscopy with machine learning to predict the extent of deterioration, allowing for early...
A new machine learning algorithm, ShotgunCSP, has been developed to predict crystal structures from material compositions with high accuracy and efficiency. This breakthrough eliminates the need for iterative first-principles calculations, making it possible to predict stable structures even for large and complex systems.
A Lehigh University team developed a novel machine learning method to predict abnormal grain growth in materials, enabling the creation of stronger, more reliable materials. The model successfully predicted abnormal grain growth in 86% of cases, with predictions made up to 20% of the material's lifetime.
MIT researchers have developed a new data-driven method that eliminates redundant computations in complex logistical problems. The approach uses machine learning to predict which operations should be recomputed and reduces the solve time for problems like scheduling trains, hospital staff, and factory tasks.
Binghamton University is launching an Institute for AI and Society with $5 million in New York state funding, enabling researchers to tackle issues like online antisemitism and protecting power systems from malicious attacks. The institute will tap into the power of Empire AI, a consortium of public and private universities in New York.
Researchers develop E2T algorithm for extrapolative predictions, outperforming conventional machine learning in material property prediction tasks. The algorithm enables models to learn broadly applicable learning methods, achieving high predictive accuracy even beyond the training data.