International researchers used machine learning to forecast marsh establishment under various environmental conditions, revealing that controllable local factors are more important than global climate change. The study suggests smart management of tidal flats can counteract threats and strengthen wetlands.
Researchers at Massachusetts General Hospital developed an AI-based method to predict atrial fibrillation risk based on electrocardiogram data. The method was highly predictive, especially in subsets of individuals with prior heart failure or stroke, and could serve as a pre-screening tool for patients at risk.
Researchers identified 166 prognostic biomarkers from long non-coding RNAs, with one biomarker, HOXA10-AS, showing high effectiveness in categorizing gliomas as low- or high-risk. The study provides potential therapeutic targets and insights into cancer biology.
Researchers at Peking University developed a non-invasive exhaled breath screening system for COVID-19, identifying key breath-borne VOC biomarkers in just 5-10 minutes. The technology reduces transmission risk by providing quick screenings for false negatives and hospital discharges.
Researchers at the University of Bern have developed an approach called 'evolving-to-learn' (E2L) that enables computers to discover mechanisms of synaptic plasticity, leading to improved learning capabilities. The algorithm was tested in three scenarios and successfully solved new tasks by mimicking biological evolution.
A new machine learning-based approach enhances student engagement in online environments. The algorithm detects when students disengage, prompting interventions to improve learning outcomes.
Researchers developed AfriBERTa, a neural network model that achieves state-of-the-art results for low-resource African languages. The model works with 11 languages spoken by over 400 million people and requires only one gigabyte of data, compared to thousands for existing models.
Researchers have developed a 'time machine' framework that uses artificial intelligence to learn from past environmental changes and predict future biodiversity loss. This framework can help decision-makers prioritize conservation approaches and mitigation interventions, leading to more effective management of ecosystem services.
Researchers have developed an autonomous robot that can open its own doors and find nearby outlets to recharge. The innovation addresses a significant challenge in robotics, enabling helper robots to work independently without human assistance.
Researchers developed a new method to predict stress at atomic scale using machine learning, enabling accurate predictions of grain boundary stresses in actual metal specimens. This breakthrough advances the field of mechanics of materials and enables scientists to engineer stronger and more heat-resistant metals.
Researchers from academia and industry will converge at Lehigh University to discuss innovative solutions for optimizing efficiency and resiliency in the global supply chain. The workshop aims to leverage machine learning for prescriptive analytics, enabling proactive optimization of supply chain operations.
A recent study published in PRX Quantum reveals that quantum machine learning algorithms are hindered by excessive entanglement, leading to a phenomenon known as barren plateaus. By limiting depth and connectivity, researchers propose a solution to avoid these regimes and successfully train quantum neural networks.
Researchers at Penn discovered a new suite of antimicrobial peptides, hidden within the human genome, which showed promising natural antibiotic potential. The peptides displayed antimicrobial activity against various pathogens, including E. coli and staph infection-causing bacteria.
FAU researchers custom-build multi-sensor tag combined with AI to observe goliath grouper species in the wild. The study identified 13 behaviors, including hovering, forward swimming, and vocalizations, using video footage from the tags.
Researchers used reinforcement learning to control a small particle moving in a double-well system, achieving accurate control despite noisy measurements. The method shows promise for future applications in quantum technologies and AI.
The Gutenberg Gait Database provides a reference set of data for healthy individuals to diagnose and treat gait disorders. The database, compiled from 350 volunteers aged 11-64, offers processed raw data and ready-to-use data for orthopedic institutes and research organizations.
A new study explores the problem of shortcuts in a popular machine learning method and proposes a solution that can prevent shortcuts by forcing the model to use more data. By removing simpler characteristics and asking the model to solve the task two ways, researchers reduce the tendency for shortcut solutions and boost performance.
A new machine learning-based algorithm has been developed to identify adolescents who have experienced suicidal thoughts and behavior. The algorithm, applied to a large dataset of survey responses from over 179,000 high school students in Utah, shows high accuracy in predicting individual adolescents at risk.
SourcePLOS·JournalPLOS ONE·TypeComputational simulation/modeling·DateNov 3, 2021
Researchers at University of Missouri and University of Chicago develop an artificial material that can respond to its environment, make decisions, and perform actions not directed by humans. The material uses a computer chip to control information processing and convert energy into mechanical energy.
Research reveals brief DBS exposure triggers significant brain state changes and sustained antidepressant response. A decrease in beta power is identified as a novel biomarker for DBS treatment optimization.
A new study improves AI diagnoses by penalizing algorithms for false negatives, which can be more urgent than accuracy. Researchers achieved significant improvements in precision and recall for chronic kidney disease and other conditions using cost sensitivity techniques.
Astrocytes play a crucial role in self-repairing the brain and may hold the key to creating energy-efficient artificial intelligence. By emulating astrocyte functions in hardware devices, researchers aim to reduce power consumption and increase fault resilience.
A new algorithm has been developed to train spiking neural networks, mimicking the human brain's structure and function. This approach enables these powerful, fast, and energy-efficient systems to solve complex tasks like image classification with high precision.
Pasqal has published a paper in the APS Physics journal presenting a new machine learning protocol called Quantum Evolution Kernel (QEK) for measuring similarity between graph-structured data on quantum computers. QEK is stable against detection error and comparable to state-of-the-art graph kernels on classical systems.
A new visual analytics tool, Sibyl, was developed to help child welfare specialists understand machine learning predictions. The tool uses bar graphs to show how specific factors of a case contribute to the predicted risk that a child will be removed from their home within two years.
A team of researchers has developed an ensemble-based machine learning model that can predict how cancer patients will respond to certain drugs with high accuracy. The model was trained on data from 499 independent cell lines and validated against a clinical dataset containing seven chemotherapeutic drugs.
Researchers at Bar-Ilan University have developed a novel treatment method that destroys cancer cells by targeting the cytoskeletal protein WASp, which is unique in active hematologic cancer cells. The approach uses small molecule compounds identified through AI and machine learning to inhibit proliferation and destroy malignant cells.
Researchers developed an automated and accurate interpretation of chest CT scans using Machine Learning technique Multiple Instance Learning (MIL). The new framework, DA-CMIL, differentiates between COVID-19 and bacterial pneumonia with performance on par to state-of-the-art methods.
Researchers at Duke University used new tools to monitor neurons and analyze machine learning data to see how zebra finches practice their courtship calls. They found that a neurotransmitter called noradrenaline shuts down variability in the song, making it more precise when performed under pressure.
Researchers at University of Pittsburgh and Prairie View A&M University developed an algorithm to repurpose cancer drugs for pulmonary hypertension, a devastating lung disease. Two compounds improved human cells and rodent markers, supporting broader drug-repurposing platform use.
A recent study published in Nature Machine Intelligence challenges the long-held assumption that accuracy and fairness are mutually exclusive in machine learning. Researchers found that optimizing models for accuracy does not necessarily compromise fairness, particularly when adjustments are made to data, labels, and scoring systems.
Research at RMIT University uses Getty's top lists of editorial pictures to analyze daily investor sentiment, predicting stock market returns based on global mood. The algorithm produces a daily score from 10 popular photos, providing a quick snapshot of investment mood across developed and emerging economies.
Researchers at Osaka University developed a deep neural network to accurately determine qubit states despite environmental noise. The novel approach may lead to more robust and practical quantum computing systems.
Researchers used machine learning on UK Biobank data to create proxy measures for brain age, intelligence, and neuroticism traits. These indirect measurements strongly correlate with specific diseases or outcomes, offering a potential solution for mental health diagnoses.
Researchers at MIT develop a data-driven process using machine learning to optimize new 3D printing materials with multiple characteristics. The system lowers costs and lessens environmental impact by reducing chemical waste and suggesting unique chemical formulations that human intuition might miss.
The INRS team has developed an intelligent optical chip that uses autonomous learning approaches to generate optical waveforms, paving the way for further advances in telecommunications. The device can autonomously adjust to a user-defined target waveform with strikingly low technical and computational requirements.
Researchers at Johns Hopkins University have developed a non-invasive optical probe to understand the complex changes in tumors after immunotherapy. Using Raman spectroscopy and machine learning, they identified key features that indicate how tumors respond to treatment, showing promising results for predicting patient response.
A new study by MIT researchers has found that blind and sighted readers have sharply different takes on what content is most useful to include in a chart caption. The study created a four-level framework for evaluating charts, which could help develop more effective tools for automatically generating captions and alternative text.
A recent study published in PLOS Medicine found that people with mental illnesses have poorer sleep quality compared to the general population. The study analyzed data from 89,205 participants and discovered significant differences in sleep patterns, including increased waking frequency and duration.
Researchers have proved the existence of universal adversarial examples that can deceive multiple quantum classifiers. The study also reveals the universality aspect of adversarial attacks for quantum machine learning systems, providing valuable insights for future applications.
Functionalized metal-organic frameworks (MOFs) show improved hydrogen interaction, increasing storage capabilities by 15-80%. The study uses machine learning to predict binding energy and reduce computationally heavy calculations.
A new algorithm developed by Carnegie Mellon University researchers offers a powerful tool for illustrating genome folding in cell nuclei. The Higashi algorithm analyzes chromatin interactions using single-cell Hi-C technology, revealing detailed variations in genome organization from cell to cell.
Researchers developed an algorithm that leverages medical informatics to predict autism spectrum disorder (ASD) diagnoses in young children. The new approach uses diagnostic codes from past doctor's visits to calculate a risk score, identifying which patients are at risk of receiving a confirmed ASD diagnosis.
A team of researchers from the University of Illinois Urbana-Champaign used advanced machine learning to model the physico-chemical properties of a molten salt compound called FLiNaK, enabling accurate atomic-scale reproduction and prediction of behavior under specific reactor conditions. This computational framework can help character...
A novel machine learning approach has been developed to understand symmetry and trends in materials, enabling researchers to group similar classes of material together. The technique uses a large, unstructured dataset gleaned from 25,000 images to identify structural similarities and trends.
A new deep-learning algorithm, ECNet, has been developed to accelerate protein engineering by predicting the fitness of all possible sequences. By incorporating evolutionary history, ECNet outperforms current methods on several datasets and identifies novel mutants with improved fitness.
Researchers developed a novel machine learning algorithm to identify previously unknown air pollutant mixtures linked to poor asthma outcomes in children. The study found that early exposure to individual and mixed pollutants can lead to longer-term problems with asthma, affecting about seven percent of US children.
A new machine learning algorithm has enabled researchers to automatically identify and map the inner structures of cells, including organelles, with unprecedented precision. By processing tens of thousands of high-resolution images, scientists have gained insights into how these structures interact and are arranged within the cell.
The University of Houston is part of a $50 million NIH-funded consortium to increase AI diversity and address health inequities. The AIM-AHEAD program will bring together experts in AI, data science, and health equity research to develop more inclusive AI tools.
Researchers at MIT develop RFusion, a robotic system that uses data from a camera and radio frequency antenna to locate and retrieve lost items. The system relies on RFID tags and machine learning algorithms to optimize the robot's trajectory and grasp the object.
Researchers at GlaxoSmithKline and CCDC combined proprietary and published datasets to train machine learning models for predicting stable polymorphs in new drug candidates. The approach leverages the large volume and variety of data in the Cambridge Structural Database, resulting in more confident predictions and improved model accuracy.
Researchers developed a dynamic respirator that modulates pore size in response to changing conditions like exercise and air pollution. The device features an AI-powered system that adjusts filtration characteristics wirelessly, providing improved breathability and comfort.
A study published in PLOS Biology suggests that machine learning models using viral genomes can predict the likelihood of an animal-infecting virus infecting humans. The researchers identified generalizable features in viral genomes that are independent of taxonomic relationships and developed models to identify candidate zoonoses.
SourcePLOS·JournalPLOS Biology·TypeComputational simulation/modeling·DateSep 28, 2021
Researchers at the University of Rochester have developed a time-domain single-pixel imaging technique that detects ultrafast light pulses with high accuracy and speed. The new method can capture 5 femtojoule pulses with temporal sampling sizes as low as 16 femtoseconds, outperforming existing methods.
A new study using machine learning uncovers 'genes of importance' in plants that help them grow more efficiently with less fertilizer, reducing economic and environmental costs. The approach also predicts additional traits in plants and disease outcomes in animals.
Researchers developed a method to apply deep learning to polarization-sensitive optical coherence tomography, enhancing cancer diagnosis. The technique enables OCT systems to detect abnormalities on a deeper level, differentiating microstructural features such as collagen fiber orientations.
Researchers developed an AI model to classify patients with localized breast cancer as high or low risk of metastatic relapse. The study shows that AI can accurately assess relapse risk with an AUC of 81%, providing a valuable aid for therapeutic decisions and avoiding unnecessary chemotherapy.
Researchers at Osaka University used machine learning to analyze locomotion data from diverse species, revealing common features associated with dopamine deficiency. The study found that worms, mice, and humans exhibit similar movement disorders when lacking dopamine, despite their evolutionary differences.
A set of guidelines published in Nature Methods provide recommendations for better reporting standards in AI methods used to classify biomedical data. The guidelines aim to ensure the quality and reproducibility of predictive methods, addressing issues such as accuracy, bias, and reproducibility.
Researchers develop crowd-assisted deep learning system to analyze disasters, integrating human intelligence with AI models for better results. The project aims to improve AI's interpretability and accuracy in disaster assessment applications.