Articles tagged with Neural Networks
Researchers at Karolinska Institutet have mapped brain networks that control the habenula, a structure linked to feelings of discomfort and aversion. The study suggests a specific pathway that can be modulated using optogenetics, offering hope for developing new treatments for depression and anxiety disorders.
A perception system for soft robots has been developed, mimicking human body components to predict complex motions and forces. The system uses a motion capture system, neural network, and soft sensors to interpret sensor signals, enabling accurate predictions of robot movements.
Researchers at Lobachevsky University have discovered the TrkB receptor system's role in forming neural networks. Activating this system increases complex functionally active neural networks with high transmission efficiency of nerve impulses, essential for learning and memory.
Researchers challenged a popular idea about how machine learning algorithms think by applying information theory to classification problems. They found that classifiers with many layers do not necessarily trade off between prediction and compression as previously thought.
Researchers at Max Planck Florida Institute for Neuroscience developed a new method to identify functional properties of individual synapses linking the two hemispheres. They found that callosal inputs and local inputs with similar orientation preference are clustered within the dendritic field, enabling coordinated network activity.
A new software, idtracker.ai, can identify up to 150 individual fish with high accuracy, extracting valuable data for understanding group behavior. The AI-powered system uses deep learning neural networks and conventional algorithms to recognize unique features of each zebrafish.
A new study led by Salk Institute researchers challenges the long-held view that individual brain cells operate as filters. The study found that the same neurons can prefer coarse or fine details depending on the context, and teaming up endows networks of neurons with flexibility to adapt to changing conditions.
Galanin-like peptide (GALP) plays a key role in regulating feeding behavior and energy metabolism. Recent research has shown that GALP administration leads to decreased respiratory quotient, indicating accelerated lipid metabolism.
Researchers developed LEAP, a neural network-based system that tracks individual body parts in video footage, enabling the study of animal behavior and neural processes. The tool has great potential outside of neuroscience, automating labor-intensive analysis to study diverse animal locomotor behaviors.
Researchers have developed a neural network that can accurately detect age and gender from video frames, outperforming existing convolutional neural networks. The new method aggregates confidence levels using mathematical statistics and Dempster-Shafer theory, enabling the detection of age and gender with improved accuracy.
A team of researchers developed a neuroinspired hardware-software co-design approach that can make neural network training more energy-efficient and faster. The approach uses a type of energy-efficient neural network called spiking neural networks, combined with the soft-pruning algorithm to minimize computing power and time.
The study combines synfire communication, coherence and resonance to provide insight into how messages are exchanged between brain areas. The researchers found that oscillations play a significant role in determining whether communication can take place.
Researchers found that isolating symptoms to a single brain area resulted in low reproducibility rates, but analyzing symptom-specific circuitry within brain networks led to 100% reproducibility. This new approach aims to shed light on therapy development by identifying common symptom-localization patterns across different diseases.
Researchers at MIT used a deep neural network to reconstruct transparent objects from low-exposure images taken in the dark. The technique could illuminate features of biological tissues and cells in low-light conditions, reducing the need for excessive light exposure that can damage specimens.
Researchers at Ruhr-University Bochum discovered that kainate receptors affect the development of brain cells immediately after birth, causing increased activity and dendrite growth. The discovery sheds light on the role of glutamate receptors in early maturation of nerve cells.
A novel framework combines CNNs and geometric inference methods to compute sketches and their corresponding 3D shapes faster and more intuitively than existing methods. Novice users can create stylish contents for their 3D creations using the tool.
Researchers develop novel method to realistically simulate dressing tasks using machine learning techniques, incorporating sense of touch to overcome challenges in cloth simulation. The approach enables single dressing sequences and a character controller that can successfully dress under various conditions.
Researchers at Carnegie Mellon University have developed a new method that uses neural networks to analyze single cell RNA sequencing data, identifying key genes and cell subtypes. This approach enables the analysis of all cell types, providing a method for comparative analysis.
Researchers have discovered three types of attacks that can be launched using a computer's graphics processing unit (GPU) to spy on web activity, steal passwords, and break into cloud-based applications. The attacks work by exploiting vulnerabilities in the GPU's memory utilization and performance counters.
A novel unsupervised language translation model developed by MIT researchers can translate far more languages with greater speed and efficiency. The model uses Gromov-Wasserstein distance to align words in two languages, achieving accurate results without human annotations.
Researchers used calcium imaging techniques to visualize spontaneous activity patterns in the mature visual cortex, finding precise organizational networks in the cerebral cortex much earlier than previously thought. These networks predict future brain function and are critical for processing complex sensory input.
Researchers discovered robust long-range patterns of correlated spontaneous activity in immature ferrets, contradicting expectations. These early activity patterns served as a template for the development of mature distributed networks, suggesting that 'local connections build a network activity scaffold'.
Scientists discovered that developing brain networks act locally to build globally, with spontaneous activity patterns correlating between distant populations of neurons. This finding suggests that long-range order originates from neural activity driven by short-range connections.
A neural network mimics the fruit fly's visual system and distinguishes between individuals based on sight alone. The system achieved an F1 score of 0.75, surpassing human biologists' performance in a similar task.
A deep neural network trained on annotated radiographs from senior orthopedic surgeons helps reduce misinterpretation of fractures in emergency department clinicians. The system demonstrated the transfer of expertise from specialists to generalist clinicians, leading to improved accuracy.
Researchers at the University of Alberta and Quantum Silicon Inc. have developed an atomic ultra-efficient electronics technology, enabling bespoke atomic patterns to control electrons. This innovation simulates neural networks, potentially training AI models more rapidly and accurately.
Researchers used deep convolutional neural networks to discriminate between signal and background tracks in the PandaX-III experiment, improving detection efficiency by 62% compared to traditional methods. The technique enhances our understanding of neutrinos and their role in matter-antimatter asymmetry.
Researchers use artificial neural networks to predict crystal stability in garnets and perovskites, achieving accuracy up to 10 times that of previous models. The team's web application allows for fast computation of material properties on various devices.
Researchers developed an artificial neural network that can compare medical concepts to specific symptoms mentioned by patients on social media. The AI system uses semantic vector word representation and can identify symptoms like insomnia or vertigo with high accuracy.
Researchers at UNIGE successfully resynchronized neurons to correct desynchronization in neural networks, suppressing behavioral symptoms associated with schizophrenia. The study, published in Nature Neuroscience, offers promising results for a new therapeutic approach targeting defective inhibitory neurons.
A team of scientists from Lehigh University has successfully engineered a living neural network that can perform basic learning tasks. The project, supported by the National Science Foundation, aims to develop new ways to think about computer design and may influence brain-related research.
Scientists at Argonne National Laboratory have developed a new method using neural networks to identify the structural signatures of molecular gases. This breakthrough enables researchers to accurately sense unidentified chemicals or scan samples for impurities in a much smaller period of time.
Scientists from TUM discovered that individual nerve cells create parallel connections to three areas of the brain, establishing feedback loops that reinforce salient stimuli while suppressing others. This automatic attention control mechanism is also shared by humans, revealing insights into perception and consciousness.
A new theory by Max Ortiz Catalan suggests that phantom limb pain results from neural entanglement between the missing limb's circuitry and pain perception networks. The 'stochastic entanglement' hypothesis explains how a novel treatment, Phantom Motor Execution (PME), can help alleviate pain by reactivating dormant brain areas.
Researchers at the University of Waterloo have developed a new AI-powered system that enables TVs to understand voice queries more accurately. The system, which was tested on Comcast's Xfinity X1 platform, can handle complex queries and personalize results based on user context.
A novel encryption technique combining homomorphic encryption and garbled circuits secures data used in online neural networks without significantly slowing their runtimes. This approach holds promise for using cloud-based neural networks for medical-image analysis and other applications that use sensitive data.
Researchers use deep neural networks to recognize images transmitted over optical fibers, achieving high accuracy despite distortions caused by environmental factors. The technique has potential for improving endoscopic imaging in medical diagnosis and increasing the information-carrying capacity of fiber-optic telecommunication networks.
Researchers found that brain activity patterns shift towards stored representations of clear images, suggesting that past experiences play a significant role in perception. The study used fMRI to analyze how the brain processes blurred images and found that higher-order circuits were more affected by clear image-induced shifts.
Researchers at NIST have developed a silicon chip that uses light instead of electricity to precisely distribute optical signals across a miniature brain-like grid. The chip enables complex routing schemes necessary to mimic neural systems and has demonstrated uniform output with low error rates.
The Network for Excellence in Neuroscience Clinical Trials has been renewed for five more years, enabling the study of new treatments for brain disorders. Nine clinical trials are currently underway, demonstrating the potential of NeuroNEXT to expedite research and bring treatments to patients faster.
Researchers found that experienced animals form memories using different plasticity mechanisms than naive subjects, suggesting the way our neurons form new connections depends on their prior history. Previously activated neurons were more excitable, making them capable of different kinds of plasticity.
Researchers trained a machine learning algorithm to analyze microscopic radiation damage, achieving an accuracy of 86% compared to humans. The algorithm can process images faster and more efficiently than humans, making it a promising tool for developing safe nuclear materials.
Artificial neural networks can now be trained directly on an optical chip, paving the way for less expensive, faster, and more energy-efficient AI. This breakthrough enables complex tasks like speech or image recognition to be performed more efficiently.
Scientists developed a neural network device using nanomaterials, generating spontaneous spikes similar to nerve impulses of neurons. The researchers replicated brain function by utilizing molecular junctions and negative differential resistance.
Researchers at UC San Diego have found that axon geometry is crucial in information flow, with a 'refraction ratio' of 0.92 indicating optimal balance between signal latency and refractory period. This discovery has implications for understanding neurological disorders like autism and developing more brain-like artificial neural networks.
Researchers at the Higher School of Economics have developed a new method for recognizing people on video using only one photo, achieving higher recognition accuracy compared to existing methods. The algorithm uses information on how reference photos are related to correct errors in video frame recognition.
Researchers at Caltech developed an artificial neural network made of DNA that can accurately identify handwritten numbers. The network, designed by Kevin Cherry, uses a 'winner take all' competitive strategy and undergoes complex reactions to classify molecular information.
Researchers developed a means of tracking retinal neuron activity as it delivers visual information to the thalamus, revealing organized clusters and shared sensitivities among different types of neurons. This finding suggests the retina's version of Pointillism, where nearby dots fuse together to create diverse colors.
Researchers analyzed data from three major brain banks and found that human herpesvirus DNA and RNA were more abundant in the brains of Alzheimer's patients. The study suggests that viruses may be involved in regulating genes associated with increased Alzheimer's risk, and could offer potential new paths for treatment.
Scientists at PNNL have developed a deep neural network that accurately detects nuclear events with high accuracy, often exceeding human expert's performance. The network was trained on 32,000 pulses and achieved impressive results, correctly identifying 99.9% of signals with minimal noise.
Researchers at MIT have developed an AI-based method to design multilayered nanoparticles with desired properties, potentially speeding up the development of new materials. The technique uses computational neural networks to learn how a nanoparticle's structure affects its behavior, allowing for faster prediction and design.
Scientists have developed a neural network that can recognize features in x-ray absorption spectra sensitive to atomic arrangement at fine scales. This method helps reveal details of atomic-scale rearrangements during iron's phase transition, and could be applied to study nanoparticles, catalytic materials, and other materials.
Researchers at The University of Tokyo Institute of Industrial Science describe a new method for creating one mini neuron network model, using microscopic plates to connect neurons together one cell at a time. This approach guides neurons to grow in a defined way and form functional communication hubs.
Houston Methodist researchers developed a lab-on-a-chip technology that models human neural networks to study retinal diseases and potential treatments. The NN-Chip can quickly screen drugs for damaged neuron and retinal connections, offering new hope for treating conditions like macular degeneration.
A groundbreaking study by Gladstone and Google AI uses deep learning to analyze cell images, identifying features that humans can't detect. The method uncovers important information that was previously impossible or problematic for scientists to obtain.
Bartos' project will examine the functional role of inhibitory nerve cells in forming memory traces and controlling cognitive behavior, a process not yet fully understood.
A smartwatch coupled with a machine learning algorithm detected atrial fibrillation (AF) with high accuracy in patients undergoing treatment for AF. The study used data from 9,750 participants and found promising results for the use of commercially available smartwatches to detect AF.
A research team led by Eiji Watanabe reproduced illusory motion using deep neural networks trained for prediction. The DNNs accurately predicted motion in unlearned videos and represented rotational motion in illusion images, similar to human visual perception.
ORNL researchers design a novel method for energy-efficient deep neural networks, achieving nearly the same accuracy as original DNNs while consuming 38 times less energy. The approach uses 'deep spiking' neural networks with stochastic-based implementation, which overcomes tradeoff between energy efficiency and task performance.
A new algorithm enables larger parts of the human brain to be represented using the same amount of computer memory, significantly reducing the memory required for simulations. This breakthrough allows researchers to simulate neuronal networks on the scale of the human brain for the first time, enabling studies of complex brain functions.