A neural network is a computational model inspired by the structure and function of the human brain. It consists of interconnected artificial neurons that process and transmit information, enabling machine learning tasks such as pattern recognition, classification, and regression by learning from labeled data.
Researchers evaluated deep learning models for waste classification in smart cities, with ResNeXt-101 emerging as the top performer. The study suggests a federated learning framework to enhance trash detection across diverse environments, leveraging multiple CNN models for improved efficiency in waste management.
Researchers in Nature Communications introduced SchNet4AIM, a model integrating SchNet for molecule interpretation with local quantum descriptors. This approach accurately predicts atomic charges and interaction energies while maintaining computational efficiency and interpretability, offering insights into complex chemical phenomena.
In their Agronomy journal article, researchers developed a method using RGB-D images and the YOLO-banana neural network to non-destructively localize and estimate the weight of banana bunches in commercial orchards.
Researchers developed an advanced automated system for early plant disease detection using an ensemble of deep-learning models, achieving superior accuracy on the PlantVillage dataset. The study introduced novel image processing and data balancing techniques, significantly enhancing model performance and demonstrating the system's potential for real-world agricultural applications.
Researchers introduced biSAMNet, a cutting-edge model integrating word embedding and deep neural networks, for classifying vessel trajectories. Tested in the Taiwan Strait, it significantly outperformed other models, enhancing maritime safety and traffic management.
Researchers from China have integrated computer vision (CV) and LiDAR technologies to improve the safety and efficiency of autonomous navigation in port channels. This innovative approach utilizes advanced path-planning and collision prediction algorithms to create a comprehensive perception of the port environment, significantly enhancing navigation safety and reducing collision risks.
Researchers developed and validated machine learning models for predicting turbulent combustion speed in hydrogen-natural gas spark ignition engines, showcasing their superiority over traditional methods. By leveraging data from a MINSEL 380 engine and employing techniques like random forest and artificial neural networks, the study demonstrated high forecasting accuracy, making these models valuable for industrial applications such as engine monitoring and simulation tools.
Researchers utilized machine learning algorithms to predict life satisfaction with high accuracy (93.80%) using data from a Danish government survey. By identifying 27 key questions and employing models such as KNN, SVM, and Bayesian networks, the study highlighted the significant impact of health conditions on life satisfaction and made the best predictive model publicly available.
Researchers in Nature explore the application of deep learning to analyze plasma plume dynamics in pulsed laser deposition (PLD). Using ICCD image sequences, a (2 + 1)D convolutional neural network correlates plume behavior with deposition conditions, enabling real-time monitoring and predictive insights for optimizing thin film growth.
DeepCNT-22, a machine learning force field, powers simulations revealing the atomic-level dynamics of SWCNT formation. It challenges conventional growth models, highlighting stochastic defects and conditions for defect-free growth.
Researchers demonstrated a novel approach to structural health monitoring (SHM) in seismic contexts, combining self-sensing concrete beams, vision-based crack assessment, and AI-based prediction models. The study showed that electrical impedance measurements and the AI-based Prophet model significantly improved the accuracy of load and crack predictions, offering a robust solution for real-time SHM and early warning systems.
Researchers combined density functional theory (DFT) with machine learning (ML) to screen 41,400 metal halide perovskites (MHPs), identifying 10 promising candidates with improved stability and optoelectronic properties. Highlighting CsGe0.3125Sn0.6875I3 and CsGe0.0625Pb0.3125Sn0.625Br3, this study offers a new framework for optimizing perovskites for solar cells.
Researchers developed advanced deep learning (DL)-based automatic feature recognition (AFR) methods that significantly enhance computer-aided design (CAD), process planning (CAPP), and manufacturing (CAM) integration. Their approach, using the multidimensional attributed face-edge graph (maFEG) and Sheet-metalNet, a graph neural network, improves recognition accuracy and adapts to evolving datasets, addressing limitations of traditional and voxelized representations.
Researchers developed a deep learning and particle swarm optimization (PSO) based system to enhance obstacle recognition and avoidance for inspection robots in power plants. This system, featuring a convolutional recurrent neural network (CRNN) for obstacle recognition and an artificial potential field method (APFM) based PSO algorithm for path planning, significantly improves accuracy and efficiency.
Researchers presented a novel dual-branch selective attention capsule network (DBSACaps) for detecting kiwifruit soft rot using hyperspectral images. This approach, detailed in Nature, separates spectral and spatial feature extraction, then fuses them with an attention mechanism, achieving a remarkable 97.08% accuracy.
Researchers present a groundbreaking holographic system in Nature, merging metasurface gratings, compact waveguides, and AI-driven holography algorithms to create vibrant 3D AR experiences. Their prototype, integrating a metasurface waveguide and phase-only SLM, achieves unmatched visual quality and represents a significant leap in wearable AR device development.
This study proposes an innovative method for detecting cracks in train rivets using fluorescent magnetic particle detection (FMPFD) and instance segmentation, achieving high accuracy and recall. By enhancing the YOLOv5 algorithm and developing a single coil non-contact magnetization device, the researchers achieved significant improvements in crack detection.
Researchers introduced a groundbreaking silent speech interface (SSI) leveraging few-layer graphene (FLG) strain sensing technology and AI-based self-adaptation. Embedded into a biocompatible smart choker, the sensor achieved high accuracy and computational efficiency, revolutionizing communication in challenging environments.
Researchers utilized long-short-term memory (LSTM) neural networks to address sensor maintenance issues in structural monitoring systems, particularly during grid structure jacking construction. Their LSTM-based approach effectively recovered missing stress data by analyzing data autocorrelation and spatial correlations, showcasing superior accuracy compared to traditional methods.
Researchers proposed a novel approach integrating machine learning with mixture differential cryptanalysis for block cipher analysis. By developing an eight-round mixture differential neural network (MDNN) and executing key recovery attacks on SIMON32/64, they showcased the method's effectiveness in enhancing accuracy and robustness in cryptographic analysis.
Terms
While we only use edited and approved content for Azthena
answers, it may on occasions provide incorrect responses.
Please confirm any data provided with the related suppliers or
authors. We do not provide medical advice, if you search for
medical information you must always consult a medical
professional before acting on any information provided.
Your questions, but not your email details will be shared with
OpenAI and retained for 30 days in accordance with their
privacy principles.
Please do not ask questions that use sensitive or confidential
information.
Read the full Terms & Conditions.