Articles tagged with Artificial Neural Networks
Researchers at KAUST developed a brain-on-a-chip that can learn real-world data patterns without extensive training, leveraging spiking neural networks and spike-timing-dependent plasticity model. The system is more than 20 times faster and 200 times more energy efficient than other neural network platforms.
A recent study found that artificial neural networks develop spontaneous systems to name colours, adopting similar behaviours to humans. The researchers identified a universal property of optimizing complexity/accuracy trade-offs in discrete communication systems.
Researchers at UC San Diego have developed a nanoscale artificial neuron device that efficiently carries out activation functions in hardware, reducing computing power and circuitry. The device, which implements the rectified linear unit activation function, can process images and perform edge detection with high accuracy.
A team of RIT researchers is working on developing an artificial intelligence system that can learn over time and play the popular video game Starcraft II. This project has the potential to advance practical solutions such as self-driving cars, service robots, and other real-world applications.
A team of researchers from UNIST developed a new learning method for PCM-based memristor neural networks, improving their learning ability by about 3% in handwriting classification tasks. The approach leverages the 'resistance drift' property of phase-change memory to update synapses and associate patterns with data.
A team of Skoltech researchers demonstrates that universal adversarial perturbations (UAPs) can be explained by classical Turing patterns. This finding can help construct a theory of adversarial examples and design defenses against pattern recognition systems.
Scientists have found a way to reduce energy consumption in deep neural networks, paving the way for more efficient AI hardware. The approach uses simple electrical impulses instead of complex numerical values, maintaining high accuracy.
Researchers developed an AI model that uses ECG data to predict atrial fibrillation and AF-related stroke risk. The model identified high-risk patients with 62% accuracy, allowing for earlier intervention.
Researchers found similar properties between deep neural networks and primate visual cortices, providing insights into attention mechanisms. The study may accelerate AI development by understanding the neural basis of attention.
Researchers from RUDN University found a way to reduce the size of a trained neural network by six times without retraining, achieving significant storage volume reduction and minimal accuracy loss. The new method leverages correlations between initial and simplified weights, eliminating the need for post-training.
A team of researchers has developed an optical computing core prototype using phase-change material, accelerating neural networks and reducing energy consumption for AI applications. The technology is scalable and directly applicable to cloud computing, making it a promising solution for the growing demands of AI online.
A breakthrough in optical neural networks accelerates computing speed and processing power to over 1000 times that of previous processors. The system can process record-sized images and achieve full facial image recognition.
Artificial neural networks enhance signal-to-background ratio in near-infrared imaging, sharpening blurred images. The technology has potential to improve diagnostics and image-guided surgery in the clinic.
A Convoluted Neural Network was trained on a library of mosquito images to classify species, sex, and strain. The system achieved 99.96% prediction accuracy for class identification.
Researchers developed an AI that can detect the type of post-stroke depression (PSD) in patients, which is crucial for providing the right treatment. The study used a probabilistic artificial neural network to analyze data from 274 patients and found that early detection is key to improving functional recovery.
Researchers exploring the nature of AI failures reveal 'adversarial examples' may not be intentional mistakes. Instead, they might be 'artifacts' created by interactions between network and data patterns. This rethink suggests that misfires could offer useful information if interpreted correctly.
Researchers studied how convolutional neural networks respond to brightness and color visual illusions, finding that they are similarly deceived as humans. The study highlights the limitations of CNNs in mimicking human vision, revealing both similarities and differences between the two.
A new data processing module called attentive normalization improves the performance of deep neural networks by combining feature normalization and feature attention. The hybrid module significantly increases accuracy while using negligible extra computational power, and facilitates better transfer learning between different domains.
A new tool, DeepTrust, generated automatic indicators of data and prediction trustworthiness in neural networks, addressing the need for trust in AI. The researchers used subjective logic to assess neural network architectures, providing insights into testing reliability and maximizing accuracy.
Researchers developed a neural network to distinguish between Middle and Late Stone Age assemblages by analyzing frequent tool combinations. The study found that the combined occurrence of backed pieces, blade technologies, and absence of core tools reliably identifies Late Stone Age assemblages.
A team of researchers at Tokyo Tech successfully used machine learning with an artificial neural network model to predict two key properties of self-assembled monolayers, enabling advanced material screening and design. This approach opens up new possibilities for the development of biomaterials with desired functions.
Researchers create novel method using artificial intelligence to connect static and dynamic calculations, enhancing system security and safety requirements. The approach enables operators to anticipate disruptions and optimize resource allocation for a more resilient power grid.
Researchers from North Carolina State University discovered that incorporating Hamiltonian function into neural networks enables them to better predict and respond to chaos. This innovation has significant implications for improved artificial intelligence applications.
Artificial neural networks became unstable after continuous unsupervised learning, but exposure to Gaussian noise mimics slow-wave sleep stabilized them. This finding has implications for the development of biologically realistic AI systems.
Scientists at UTokyo-IIS developed a machine learning algorithm to infer excited states from ground states of materials. The algorithm used artificial neural networks to analyze data from core-electron absorption spectroscopy, revealing new insights into chemical reactivity and material function.
Researchers developed a novel AI-managed trading strategy that outperforms traditional methods, achieving greater gains and fewer losses. The proposed system utilizes convolutional neural networks to analyze layered images of current and past market data, leading to more accurate predictions and reduced randomness.
Researchers developed Early Bird, an energy-efficient method for training deep neural networks, which can use 10.7 times less energy than traditional methods to achieve the same level of accuracy. This breakthrough could lead to significant cost savings and a reduction in greenhouse gas emissions.
A new chip has been developed at TU Wien that can recognize certain objects within nanoseconds, leveraging artificial intelligence and a special material. The chip integrates the neural network with its AI directly into the image sensor, making object recognition faster by many orders of magnitude.
Insects can perform numerical estimation and compare sets of objects, recognizing larger quantities. Researchers found a simple model for neural networks to learn numerical cognition tasks using a single neuron.
A novel nanoelectronics device has enabled brain neurons and artificial neurons to communicate with each other over the internet. This breakthrough study shows how three key emerging technologies can work together: brain-computer interfaces, artificial neural networks and advanced memory technologies.
A new artificial neural network model has been developed to solve inverse problems, demonstrating accuracy comparable to the maximum entropy (MaxEnt) approach. The model's versatility and robustness against noisy data have been showcased in various tests, including recovering electron single-particle spectral densities.
Researchers at Otto-von-Guericke-Universität Magdeburg are using brain research methods to analyze artificial neural networks and improve explainable AI. The Cognitive neuroscience inspired techniques project aims to understand the internal processes of ANNs and identify malfunctions.
Researchers at Politecnico di Milano developed a novel circuit that can execute advanced AI operations in one operation, reducing energy consumption and paving the way for more sustainable AI computing accelerators. This breakthrough enables faster and more efficient training of neural networks, crucial for applications like facial rec...
Researchers have developed a custom artificial neural network that can analyze molecular signals controlling gene function, enabling biologists to understand complex mechanisms of gene regulation. This breakthrough enables the creation of machine learning algorithms that reflect common concepts in biology.
Researchers at the University of Pittsburgh have developed an artificial synapse that mimics the human brain's ability to create neuronal connections. This breakthrough technology could revolutionize AI and cognitive computing, enabling faster and more efficient processing of complex tasks.
Researchers developed a novel computational approach using deep artificial neural networks to predict neural responses to images. The study found that certain stimuli, such as checkerboards or sharp corners, elicit strong responses from neurons, contradicting current dogma in the field.
Researchers trained neural networks to predict plant growth patterns using computer vision algorithms and efficient graphics processing units. The system uses Raspberry Pi with Intel Movidius graphics card to calculate and predict the optimal ratio of nutrients, enabling continuous monitoring and prediction in artificial growing systems.
A new study reveals that human brains process faces in a similar way to artificial intelligence systems, with unique activation patterns playing a key role in recognition. The researchers found parallels between the human visual system and deep neural networks, which can improve face recognition capabilities.
Artificial neural networks successfully identify minor changes in DNA structure caused by UV radiation, enabling early detection of potential cancer risks. The technique uses surface-enhanced Raman spectroscopy and has the potential to be used for medical diagnostics.
Researchers at the University of Delaware are developing new memory devices that can support neural networks in low-power embedded systems. These advancements aim to improve the lifetime and reliability of IoT devices, which currently struggle with battery power and memory constraints.
Researchers use deep neural networks to simulate light-induced molecular reactions on long time scales, accelerating computation by up to 19 years. This method enables better understanding of biological processes like carcinogenesis and ageing, with potential applications in material ageing and photosensitive drugs.
Bengio will discuss recent successes and limitations of deep learning, as well as research directions for building human-level AI. The event is part of the Heidelberg Laureate Forum, a networking event bringing together young researchers and award recipients.
A two-layer all-optical artificial neural network has been successfully demonstrated for complex classification tasks, outperforming computer-based neural networks. The researchers plan to expand this approach to large-scale optical deep neural networks for specific practical applications.
Researchers believe that studying animal brains can improve AI's ability to tackle complex tasks like dish-washing. By understanding how biological neural networks work, AI systems may be able to overcome barriers and achieve superhuman performance.
Researchers have developed an effective method to monitor carbon nanotube films using artificial neural networks (ANN). The technique can help predict the efficiency of single-walled carbon nanotubes synthesis and improve the overall production framework, leading to new horizons for real-life applications.
Scientists from Russia and Greece have successfully implemented a spiking neural network based on memristors, demonstrating the feasibility of local learning rules. The research enables autonomous unsupervised learning of complex neural networks, paving the way for new applications in AI.
Researchers employed machine learning to analyze images of quantum systems and identify the most predictive theory. The study used artificial neural networks to distinguish between competing theories, selecting the one that best described observed phenomena in high-temperature superconductors.
Scientists have made a breakthrough in understanding the behavior of strongly interacting electrons using machine learning techniques, discovering a new state called Vestigial Nematic State. The technique uses artificial neural networks to recognize different forms of electronic matter and reveals symmetries of complex image-arrays fro...
Researchers controlled specific neurons in macaques' brains with images generated by artificial neural networks, demonstrating a new tool for neuroscientists to design experiments. This breakthrough uses current computer vision applications to predict and control visually evoked neural responses in primate brains.
A team of scientists at MIT developed a neural network that can read scientific papers and generate a plain-English summary. The system, called RUM, uses vectors rotating in multidimensional space to represent words and improve memory and recall capabilities.
Researchers developed Bright-field Holography to overcome limitations of holographic 3D imaging. The method combines the image contrast advantage of bright-field microscopy with the snapshot volumetric imaging capability of holography, allowing for rapid creation of images equivalent to those from a bright-field microscope.
A team at NYU Tandon School of Engineering has designed an artificial neural network approach that can predict the elastic modulus of graphene-enhanced composites from just one sample, streamlining materials testing. This reduces the need for extensive experimentation, lowering costs and accelerating product development.
Researchers used artificial neural networks to learn atomic interactions from quantum mechanics, bypassing complex calculations. The ANN was used as a surrogate model to account for errors and improve predictions.
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.
Researchers at NYU Tandon School of Engineering have developed a machine learning system that pairs artificial neural networks with infrared imaging to control and interpret small-scale chemical reactions. This technique can reduce the decision-making process from one year to weeks, saving tons of chemical waste and energy.
Researchers at Osaka University have developed an AI-based system to track individual fluorescently labeled molecules in living cells. The system can analyze hundreds of thousands of molecules in a short period, providing reliable data on molecule status and dynamics.
A study by Cameron Buckner using deep neural networks suggests that human knowledge stems from sensory experience, a school of thought known as empiricism. The networks demonstrate how abstract knowledge is acquired and can be used to understand complex tasks in neuroscience and psychology.
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 at MIT develop Temporal Relation Network (TRN) module to help CNNs recognize activities by observing key frames. The module achieves top accuracy of 95% in activity recognition on Jester dataset, outperforming existing models.
Using deep learning algorithms, researchers have developed a system that forecasts aftershocks significantly better than random assignment. By analyzing earthquake data and physics-based models, they identified the second invariant of the deviatoric stress tensor as an important factor in predicting aftershock locations.