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 quantum-classical approach has been developed for designing photochromic materials, accelerating the discovery of novel compounds. The method identified five promising candidates with key properties essential for photopharmacology applications.
The SETI Institute invites researchers to apply for a postdoctoral fellowship focused on developing advanced AI tools for exoplanet discovery. The successful candidate will work with Dr. Vishal Gajjar and his team to uncover subtle signals in massive datasets using machine learning and anomaly-detection techniques.
A recent study emphasizes the urgent need to address bias in generative AI systems, which can distort outcomes and erode public trust. The research suggests that developing and deploying ethical, explainable AI is crucial to ensure fairness and transparency in critical decision-making areas.
Researchers developed a machine learning tool that screens for comorbid depression and anxiety disorders using acoustic voice signals. The study confirmed that a one-minute verbal fluency test can reliably identify subjects with comorbid AD/MDD, who tend to use simpler words and exhibit reduced variability in phonemic word length.
Yann LeCun, NYU's Courant Institute of Mathematical Sciences professor, has been selected as a winner of the 2025 Queen Elizabeth Prize for Engineering for his groundbreaking research on artificial neural networks. His work enabled machines to process and learn from vast amounts of data in ways previously unimaginable.
Researchers found that a machine learning predictive model leveraging EHR data outperformed screening surveys in identifying patients with health-related social needs. The models demonstrated biases, particularly towards White, non-Hispanic patients.
Dental implant surgeries require optimal mechanical stress levels for successful bone healing and long-term implant success. Researchers are developing a hybrid biomechanical model using machine learning to provide precise, patient-specific predictions of mechanical stress.
A team of MIT engineers has developed a training method for multiagent systems that can guarantee their safe operation in crowded environments. The method enables agents to continually map their safety margins, allowing them to scale up to any number of agents while maintaining system safety.
Researchers developed SPACIER, an open-source software that integrates machine learning with molecular simulations to design high-performance optical polymers. The tool surpassed the empirical limits of refractive index and Abbe number in a proof-of-concept study, demonstrating its practical potential.
A new machine learning approach developed by UCR researchers enables the detection of environmental patterns in large datasets generated by LIGO, improving data quality and reducing noise. This technology has potential applications in particle accelerator experiments and complex industrial systems.
Dr. Girish N. Nadkarni, a pioneering physician-scientist, has been named Chair of the Windreich Department of Artificial Intelligence and Human Health and Director of the Hasso Plattner Institute for Digital Health at Mount Sinai. He will lead efforts to advance AI research, education, and clinical translation.
Researchers found that training AI agents in a less noisy environment can lead to better performance than traditional methods. The indoor training effect suggests that constructing simulated environments with specific noise levels can improve AI learning.
Researchers developed ProtET, an AI model leveraging multi-modal learning to controllably edit proteins through text-based instructions. This approach enhances functional protein design across domains like enzyme activity, stability, and antibody binding, promising real-world applicability in biomedical research.
The open-source AI model analyzes medical images, generates detailed reports, and answers clinical questions to streamline diagnostics and improve accuracy. BiomedGPT aims to democratize healthcare and reduce disparities amongst patients by providing easily accessible data to bolster underserved hospitals.
Researchers developed MUNIS, a deep learning tool that predicts CD8+ T cell epitopes with high accuracy, potentially accelerating vaccine development. The tool was validated using experimental data from influenza, HIV, and EBV, demonstrating its potential to streamline vaccine design.
A collaborative study investigates how advanced AI tools can make it easier to evaluate interventions for ageing, providing personalised recommendations. The researchers identified eight critical requirements for effective AI-based evaluations and found that following specific guidelines improved the quality of the recommendations.
Vision-Language Models (VLMs) inherit biases from uncurated datasets, leading to poor group robustness and biased predictions. Researchers are exploring strategies to mitigate these biases in discriminative models, but generative tasks like image captioning and image generation require attention.
Researchers designed nano-architected materials with exceptional strength-to-weight and stiffness-to-weight ratios, overcoming stress concentration issues. Machine learning optimized geometries led to over double the strength of existing designs.
A recent study has employed machine learning algorithms to improve the accuracy of flood season rainfall predictions. The findings show that combining climate system numerical models with ML-based correction methods results in substantial improvements, increasing prediction scores by up to 7.87%.
Neuromorphic computing is poised to emerge into full-scale commercial use, driven by the need for energy-efficient solutions. The review article proposes strategies for building large-scale neuromorphic systems that can tackle complex real-world challenges.
Researchers applied machine learning methods to analyze large datasets of individual cells, achieving better results than classical learning methods. Self-supervised learning improved performance, especially in applications with large single-cell data sets.
A new hybrid machine learning model predicts ultimate axial strength of CFRP-strengthened CFST columns with high accuracy, enabling safer and more efficient designs. The model can be used to optimize construction processes and enhance the safety of structures at a lower cost.
The new tool will enhance the DLA's supply chain management capabilities, reduce operational disruptions, and bolster readiness. Quantum Research Sciences' technology will provide predictive capabilities and automate obsolescence management processes.
A groundbreaking AI model developed by researchers at Emory University accurately predicts the likelihood of blood transfusion in non-traumatic ICU patients, addressing longstanding challenges in predicting transfusion needs. The model achieved exceptional performance metrics, including an AUROC of 0.97 and an accuracy rate of 0.93.
Researchers developed a machine learning model using hyperspectral imaging to assess pre-harvest tomato quality, predicting key parameters like weight, firmness, and lycopene content. The approach enables real-time monitoring of fruit development, improving crop quality and reducing waste.
BioChatter bridges the gap between large language models and biomedical research by providing a transparent and adaptable framework for custom research tasks. The platform can integrate with knowledge graphs and bioinformatics tools, making it easier for researchers to analyze complex datasets.
Professors Scott Acton and Mathews Jacob of UVA's Charles L. Brown Department of Electrical and Computer Engineering were named to the IEEE Signal Processing Society's 2025 Class of Distinguished Lecturers for their groundbreaking work in signal processing, artificial intelligence, and medical imaging.
A new study assesses the historical knowledge of AI chatbots like ChatGPT-4 and finds they struggle with nuanced, PhD-level inquiry. The models performed best on legal systems and social complexity but struggled with topics such as discrimination and social mobility.
An English literature graduate has developed a new method for large language models to understand and analyze short text chunks, such as those on social media profiles. The method successfully grouped nearly 40,000 Twitter user biographies from accounts tweeting about US President Donald Trump into 10 categories.
A brain-computer interface has enabled a person with tetraplegia to control a virtual quadcopter by thinking about moving their unresponsive fingers. This technology provides unprecedented control, allowing the user to maneuver through a virtual obstacle course and potentially enabling remote work and social interactions.
Researchers at University of Birmingham have discovered three new protein biomarkers TFF3, LCN2, and CEACAM5 that show strong predictive potential for colorectal cancer. These biomarkers are linked to cell adhesion and inflammation, processes closely associated with cancer development.
Zhu has made groundbreaking contributions to earthquake monitoring using deep-learning models like PhaseNet, DeepDenoiser, and GaMMA. His work has led to breakthroughs in seismic phase picking, denoising, and phase association.
Researchers at NIMS developed a next-generation AI device leveraging ion-controlled spin wave interference in magnetic materials, outperforming conventional devices by up to 10 times. The technology enables energy-efficient computations with minimal degradation when miniaturized, opening doors for various industrial applications.
Researchers at TU Graz are developing a self-learning AI system to position individual molecules quickly and autonomously, enabling the construction of highly complex molecular structures. The goal is to build logic circuits in the nanometre range using quantum corrals made from complex-shaped molecules.
Researchers introduced a novel approach to enhance reservoir computing, incorporating a generalized readout that offers improved accuracy and robustness compared to conventional methods. The new method uses a nonlinear combination of reservoir variables to uncover deeper patterns in input data.
The study utilizes infrared spectroscopy and a machine-learned protocol to map spectroscopic fingerprints to atomistic structures. The authors demonstrate the accuracy of their network in predicting local atomistic structures and energetic variations, enabling the tracking of dynamic C–C coupling on Cu surfaces.
A new automated job hazard analysis tool promises to significantly reduce workplace accidents and improve safety in the construction industry. The University of South Australia's research team has built a 'knowledge graph' to predict hazards, which can be analysed in real-time to identify potential risks and control measures.
A new study by researchers at the University of Minnesota found that the benefits of corn-soybean crop rotation are extremely sensitive to climate change. The study suggests that increasing crop rotation can improve overall yields and highlight its potential as a climate adaptation strategy in the US Midwest.
A new study uses machine learning models to identify women experiencing severe subjective cognitive decline during the menopause transition, associated with aging, hypertension, obesity, and depression. This predictive model allows for early intervention to protect cognitive health, a novel guidance for interventions designed to preser...
Researchers have found evidence of songbirds forming social connections and potentially exchanging information about their migration routes through vocalizations. The study suggests that social cues play a significant role in shaping migration behaviors, particularly for young birds learning from observing other birds.
Researchers developed an AI model to detect brain cancer spread in surrounding tissue using MRI scans, showing 85-per-cent accuracy. This non-surgical method offers insights into patients' cancer without aggressive surgery, potentially improving treatment and survival.
Temperate savannas in eastern China have been mapped for the first time, revealing their geographical distribution and extent. The research provides precise information on the spatial characteristics of these ecosystems, supporting conservation and utilization efforts.
DNNs have an inbuilt 'Occam's razor,' favouring simpler solutions that fit training data. This bias helps them generalize well on simple patterns but may struggle with complex data, aligning with real-world data characteristics.
Researchers developed a machine learning model to identify defective products in semi-solid die casting by analyzing injection pressure. The model achieved high accuracy and revealed mechanisms behind defect formation, providing a foundation for optimizing manufacturing processes.
A new AI system analyzed electronic health records of long-COVID patients to identify four sub-populations with specific needs, including those with asthma or mental health conditions. The study found that these sub-populations require more specialized care and pointed toward updated profiles for hospitals to better address their needs.
The BiliSG app uses machine learning to analyze skin color and predict bilirubin levels in newborns, offering a convenient alternative to traditional testing methods. With 100% sensitivity, the app has shown promising results in monitoring neonatal jaundice and reducing the risk of brain damage.
Researchers at the University of New Hampshire developed an AI-powered algorithm to categorize over 706 million aurora images from NASA's THEMIS data set. This labeled database can help scientists better understand and forecast geomagnetic storms that disrupt vital communications and security infrastructure.
A pioneering new mathematical model developed by Oxford researchers could help assess the risks posed by AI and protect people's privacy. The method provides a robust scientific framework for evaluating identification techniques, including browser fingerprinting.
A series of small earthquakes in Surrey in 2018 and 2019 may have been triggered by oil extraction from a nearby well, according to a new study. The research used mathematical modeling to predict the frequency of earthquakes based on oil extraction timing and volume, finding a rough match with observed seismic activity.
TabPFN learns causal relationships from synthetic data, making correct predictions more likely than existing algorithms. The model requires fewer resources and data, making it ideal for small companies and teams.
Researchers developed an AI-based method to analyze CEO depression from vocal acoustic features in conference calls. The study found that CEOs with higher levels of depression tend to receive larger compensation packages and are more responsive to negative feedback.
Positive Phase 1 trial results suggest ISM5411's gut-restrictive property and favorable pharmacokinetic profile, validating its potential for treating inflammatory bowel disease. Insilico Medicine expects to initiate a Phase 2 proof-of-concept study in active ulcerative colitis patients.
A new AI model developed by researchers at Penn State College of Medicine can predict the progression of autoimmune disease among those with preclinical symptoms up to 1,000% more accurately. The GPS model integrates data from large genetic studies and electronic health records to identify individuals at high risk of disease progression.
A new method called Annotatability helps identify mismatches in cell annotations and better characterizes biological data structures. This approach enables more precise downstream analysis of biological signals, capturing cellular communities associated with target signals.
A team of researchers at the Indian Institute of Science (IISc) has developed a machine learning-based approach to predict material properties using limited data. By leveraging transfer learning and multi-property pre-training, they were able to improve model performance and extend its applicability to new materials.
A team of researchers developed a machine learning framework to streamline the discovery of high-performance ionic thermoelectric materials. The approach predicted Seebeck coefficients with high accuracy and identified critical molecular descriptors influencing material performance.
A new study uses machine learning to reduce time needed for calculating screening parameters in Koopmans functionals, enabling faster predictions of material spectral properties. Researchers trained a simple model using modest data and achieved accurate results, paving the way for studying temperature-dependent spectral properties.
Current energy-hungry transformer-based systems contrast with Turing's idea of machines that develop intelligence naturally, like human children. AI systems can now perform tasks exclusive to human intellect, such as generating coherent text and discussing abstract ideas, but with limitations on sustainability and societal impact
Researchers developed a novel AI method using Disentangled Variational Autoencoder (D-VAE) for inverse materials design, making the process data-efficient and interpretable. The method was tested on high-entropy alloys, producing clear results that highlight influencing material features.
Researchers create SciAgents framework to autonomously generate and evaluate promising research hypotheses in biologically inspired materials. The framework uses graph reasoning methods to organize relationships between scientific concepts, mimicking biological systems.