Articles tagged with Algorithms
Researchers at Stanford University have created a deep learning algorithm that can accurately predict the toxicity of chemicals and associate drugs with side effects using just six data points. This breakthrough could help chemists choose promising candidates and accelerate drug development.
Researchers at Silicon Studio developed an algorithm that uses 'ensemble' and survival analysis methods to predict when users will leave a mobile game. The model can automatically adapt to different games and data, enabling companies to understand user needs and design more entertaining games.
Physicists have developed a new feedback controller to control fusion plasma energy and rotation. The algorithm uses sensors, algorithms, and actuators to modify the plasma's rotation profile and stored energy.
A new algorithm allows users from different network providers to pair up and make better use of the available wireless spectrum, reducing inefficiency in wireless technology. The 'blind' matching algorithm uses a simple learning process and converges to a stable-matching state, enabling mutually beneficial partnerships.
Researchers developed an algorithm that uses Deep Learning to predict the decision of hematopoietic stem cells to become a certain cell type. This enables earlier detection and analysis of blood cell development, paving the way for new treatments and insights into developmental traits.
Researchers discovered that an algorithm called additive increase, multiplicative decrease (AIMD) is used both in engineered systems like the Internet and biological networks like the human brain. This finding sheds light on how the brain manages information and potentially helps understand learning disabilities.
Scientists at the University of Luxembourg developed Equihash, a memory-hard problem algorithm that resolves Bitcoin's centralization issue. This allows for more democratic digital currencies like Zcash, where users can contribute to mining with standard hardware, reducing investment costs and increasing decentralization.
Researchers tested automated approaches to analyze large linguistic datasets, finding that some methods detected cognates correctly with high accuracy. The study suggests a promising future for combining algorithms and expert knowledge to uncover human prehistory and cultural evolution.
MIT researchers incorporated human strategies into automatic planners, achieving significant improvements in performance. By encoding high-level strategies from skilled human planners, they improved the performance of competition-winning algorithms on complex planning problems.
Researchers developed machine learning algorithms to reconstruct 3D protein structures using microscopic images, enabling faster discovery of new drugs for diseases like Alzheimer's and cancer. The approach eliminates prior knowledge requirements, making it possible to study previously inaccessible proteins.
Researchers at Lawrence Berkeley National Laboratory have developed a machine learning algorithm to predict point defects in intermetallic compounds with high accuracy. This method accelerates research on new advanced alloys and lightweight materials for various industries.
A Stanford-developed deep learning algorithm accurately diagnoses skin lesions as well as board-certified dermatologists, with potential to provide life-saving smartphone diagnosis. The algorithm was trained on a database of nearly 130,000 images and tested against 21 dermatologists in identifying malignant carcinomas and melanomas.
Researchers at MIT's Sloan Neuroeconomics Lab have developed a new method to extract correct answers from large groups of people, reducing errors in crowd wisdom surveys. The 'surprisingly popular' algorithm uses the variation between individual responses and predicted popular opinion to identify the correct answer.
Researchers at Cornell University developed an artificial Ms. Pac-Man player that achieved a laboratory score of 43,720, surpassing the existing high score for computerized play. The player uses a decision-tree approach and demonstrates accuracy in predicting ghost movements with 94.6-percent accuracy.
Researchers found that attackers can crack Android Pattern Lock within 5 attempts using video and computer vision algorithm software. Complex patterns are actually easier to crack than simple ones, making shorter, simpler patterns a more secure option.
Researchers develop a treatment algorithm for bionic hand reconstruction in patients with severe nerve damage, offering improved hand function and reduced chronic pain. The study's five patients show significant improvements in hand function, including increased sensitivity and motor control.
Researchers developed a novel-machine learning method to distinguish people with common names, focusing on relational features, text features, and venue features. The new approach is an improvement over existing methods, enabling the identification of previously unencountered individuals.
Researchers created tools called disturbance observers using fractional calculus, which estimate and eliminate disturbances in systems. The new approach outperformed existing methods, particularly when combined to handle highly fluctuating disturbance signals.
Researchers developed an algorithm that extracts diverse subsets from large data sets, improving machine learning efficiency and diversity. The new method is 1 billion times faster than existing algorithms, enabling real-time analysis of vast amounts of data.
Researchers have combined CRISPR gene editing with single-cell genomic profiling to understand nuanced cellular processes. The new technology enables precise manipulation of genes in individual cells, revealing previously unknown functions and advancing the field of genetic engineering.
The UCLA team implanted a spinal stimulator in a California man who broke his neck in a dirt-biking accident, showing early promise in returning hand strength and movement. The device bypasses the injury by training the spinal cord to find alternate pathways, allowing patients to regain mobility in their hands.
The Jefferson Lab-NVIDIA collaboration uses Titan's supercomputer to simulate QCD interactions, achieving speedups of seven- to tenfold for calculations. This enables researchers to explore exotic mesons and advance QCD theory.
Game theory and algorithms improve security by randomizing patrols to evade attackers' predictions. Real-world applications have led to measurable improvements in security for US Coast Guard ferry protection since 2011.
A new recommendation algorithm allows individuals and items to belong to multiple overlapping groups, making it more realistic than existing models. The algorithm's predicted ratings proved more accurate than those from existing systems on five large datasets.
Researchers developed a machine-learning model that can distinguish between pathological hypertrophic cardiomyopathy (HCM) and physiological changes in athletes' hearts, enabling easier diagnoses. The model demonstrated superior diagnostic ability comparable to conventional 2D echocardiographic and Doppler-derived parameters.
Researchers found a universal principle underlying human intelligence, enabling complex brain computations and generalization of knowledge. The Theory of Connectivity proposes a simple mathematical logic guiding neural connections and communications.
Researchers have developed an algorithm to detect fake reviews on ecommerce sites, analyzing behavior and content features to identify deceptive posters. The method outperforms earlier detection algorithms, providing a more accurate picture of product ratings.
Researchers have developed a new algorithm that can efficiently fit probability distributions to high-dimensional data, even when the dataset contains corrupted entries. The algorithm relies on two insights: selecting an appropriate metric for measuring distance from distributions and identifying regions where cross-sections should begin.
Researchers at Northeastern University have proposed a way to optimize power exchange between the main grid and multiple microgrids using consensus-based algorithms. These algorithms allow decentralized generators to communicate with each other and with the main grid, ensuring reliable and cost-effective energy distribution.
The UnBias project aims to establish a system of auditability and build trust in the internet by studying user experience and algorithm design. Researchers will also produce educational materials to support youth understanding about online environments.
Researchers at Sandia National Laboratories explore neural computing applications, including adaptive learning, dynamical systems, and spiking network algorithms. These approaches aim to overcome the static learning bottleneck and enable precise computations.
Researchers construct networks of organelle functional modules in Arabidopsis using a soft thresholding approach. The algorithm identifies strongly co-expressed genes and links them to infer the function of unknown genes.
A new computational imaging method identifies letters printed on the first 9 pages of a stack of paper, demonstrating the feasibility of reading closed books. The MIT system exploits the unique properties of terahertz radiation to penetrate surfaces and analyze materials in thin layers.
A new study explores the potential use of artificial intelligence to analyze complete lung function tests and improve diagnosis accuracy. The algorithm simulates complex reasoning to provide a standardized and objective diagnosis, removing bias.
Researchers at MIT have developed programmable routers that can implement diverse traffic management schemes, improving network resilience. The new design allows for flexible traffic management without compromising operating speeds, enabling innovation and rapid prototyping.
A study published in the American Journal of Roentgenology suggests that nephron-sparing treatment for small renal masses can improve life expectancy in patients with mild or moderate chronic kidney disease. The study developed a framework for incorporating tumor imaging features and renal function into treatment selection, providing a...
Researchers have developed a cost-effective method using the Microsoft Kinect to evaluate gait abnormalities in multiple sclerosis patients. The device detects movement and computer algorithms quantify walking patterns, reducing human error. The technology shows promise for diagnosing gait pathology and tracking treatment effects.
Bernd Bickel's algorithm improves technical modelling of planar-rod structures for optimal stability and efficiency in manufacturing. The software has applications in art, wire sculptures, and rapid prototyping.
Researchers at University of California - San Diego create method to improve light interaction with small details called glints on surfaces, enabling accurate rendering of materials with metallic finishes and injection-molded plastics. The new algorithm is 100 times faster than current methods and can be used in animations.
A new character animation technique has been developed by Disney Research, eliminating unsightly artifacts that occur when bending joints in computer animations. The method uses pre-computed centers of rotation to calculate skin deformation, minimizing volume losses and bulging.
Biologists have developed an algorithm predicting protein cluster structure, enabling faster understanding of cellular functions and potential treatments. The new method is up to 100 times faster than previous methods, taking just 15 minutes to run on a personal computer.
The NIST Ballistics Toolmark Research Database provides a statistical foundation for reliably linking bullets to the guns that fired them. The database uses 3D topographic surface maps to analyze bullet markings, allowing researchers to quantify uncertainty and develop more accurate bullet-matching algorithms.
Researchers from Binghamton University and Stevens Institute of Technology found that smartwatches can crack private PINs with 80-percent accuracy on the first try. The study used data from embedded sensors in wearable devices to break codes without contextual information, highlighting security vulnerabilities in these devices.
A team of researchers has developed a method to identify different subtypes of neurons in the human brain, revealing unique characteristics that can lead to differences in cellular function. The study provides a unified framework to analyze individual neurons and could help diagnose and treat brain disorders.
Researchers tested facial recognition algorithms on a dataset of one million images from around the world and found that accuracy rates dropped significantly when confronted with more distractions. Google's FaceNet performed strongest, but other algorithms struggled to maintain high accuracy rates at scale.
Researchers from MIT's CSAIL present a new chip design called Swarm that makes parallel programs run many times faster and requires one-tenth the code. The chip has extra circuitry to prioritize tasks and handle synchronization between cores, making it easier for programmers to adapt sequential algorithms.
Researchers at North Carolina State University developed a new image segmentation technique that improves object identification and separation in images. The technique, called Consensus-Based Image Segmentation via Topological Persistence, aggregates data from multiple algorithms to create a new version of the image.
A new mathematical framework developed by Robert Saye accurately resolves intricate fluid dynamics near evolving interfaces, allowing for the study of complex phenomena like bubble aeration and propeller blade optimization. High-order methods outperform low-order methods in accuracy and computing power.
A new computer algorithm enables scientists to use satellite data to determine groundwater levels across larger areas than ever before. The technique, developed at Stanford University, could lead to better models of groundwater flow and monitoring in agricultural regions. The team was able to calculate surface deformations and groundwa...
Researchers developed an algorithm that can stitch together astronomical measurements from around the globe, producing the first images of black holes. The algorithm uses a technique called interferometry and preserves continuity in the image by using a mathematical model inspired by rubber sheets and plastic wrap.
Researchers found that 60% of Amazon sellers using algorithmic pricing have prices higher than the lowest price for a product. The study also revealed that algorithmic pricing leads to significant volatility in prices, with some products changing price over 8 times per day.
Two studies found that a high-sensitivity cardiac troponin I assay can accurately diagnose or rule out heart attacks in patients with chest pain. The tests identified 19% of patients as potentially suitable for immediate discharge and had high diagnostic performance in excluding heart attacks. A new algorithm using this test showed rap...
Researchers show that Super Mario Brothers' level completion can be constructed to be PSPACE-hard, a class of problems considered even harder than NP. This suggests that completing a difficult level in Super Mario Brothers may require an exponential amount of time and computational effort.
Researchers at Binghamton University have developed an algorithm using Singular Spectrum Analysis (SSA) to quickly detect and locate power grid faults, providing a promising solution for authorities to remotely identify issues before they cause prolonged outages or equipment damage.
A clinical study confirmed a strong correlation between minor heart rate irregularities and the survival prospects of heart attack patients. The new method, measuring respiratory sinus arrhythmia, produces a specific picture of the body's functional condition.
Scientists at the Australian National University have created a new algorithm that can accurately predict tsunamis by analyzing real-time ocean sensor data. The Time Reverse Imaging Method takes data from sensors to recreate what the tsunami looked like when it was born, allowing for better predictions of its trajectory.
A novel technique called model predictive path integral control (MPPI) helps autonomous vehicles maintain stability at the edge of handling limits. By leveraging advanced algorithms and onboard computing, MPPI enables vehicles to optimize their trajectories in real-time, reducing the risk of accidents on hazardous roads.
Researchers at University of Notre Dame develop a new algorithm to more accurately represent complex dependencies in data, leading to better predictions for invasive species, shipping networks and infectious diseases. The approach can also be applied to other fields such as social media, telecom companies and urban planning.
Researchers at George Mason University are working on a new method called 'shuffling' to quickly separate hackers from legitimate users and isolate bad actors. The approach aims to prevent denial-of-service attacks that can cripple online businesses with downed websites, financial losses, and damaged client relationships.
Researchers developed a platform for autonomous aerial refueling of UAVs, utilizing binocular vision-based sensor and navigation system. The system allows the tanker UAV to capture images of the receiver UAV and estimate its pose, then control the boom towards the receptacle for refueling.