Multi-view inference (MVI), which accepts images from multiple viewpoints as input of deep neural networks, is proposed to improve the inference accuracy of conventional single-view models. However, existing mechanisms face challenges in feature fusion and computation efficiency: (1) features from inter-view and intra-view contribute differently to inference, and uniform feature fusion limits MVI accuracy; (2) the sophisticated process and tremendous computational workload of MVI cause a considerable increase in inference latency. This article addresses the above challenges and enables high-accuracy and low-latency MVI for edge intelligence by proposing an end-to-edge synergistic multi-view inference (SMVI) framework. SMVI integrates the feature fusion module based on pairwise mutual-attention (FUMA), which incorporates the differences between features, enhancing MVI accuracy. To optimize the computation of FUMA-based SMVI, we present a joint optimization algorithm of resource allocation and model partition (RAMP) to reduce MVI latency, considering device heterogeneity, dynamic network connection, and resource limitation in heterogeneous edge environments. We developed an SMVI prototype system with heterogeneous embedded GPUs and evaluated its performance in real-world MVI scenarios. Extensive experiments demonstrate that the proposed mechanism achieves a notable MVI accuracy improvement of approximately 4% and accelerates the process by 4.08 & times; compared to state-of-the-art approaches.
Accurate short- and medium-term forecasting of photovoltaic (PV) power generation is vital for grid stability and renewable energy integration. This study presents a comparative scenario-based approach using Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU) models trained with one year of real-time meteorological and production data from a 250 kWp grid-connected PV system located at Dicle University in Diyarbakır, Southeastern Anatolia, Turkey. The dataset includes hourly measurements of solar irradiance (average annual GHI 5.4 kWh/m2/day), ambient temperature, humidity, and wind speed, with missing data below 2% after preprocessing. Six forecasting scenarios were designed for different horizons (6 h to 1 month). Results indicate that the LSTM model achieved the best performance in short-term scenarios, reaching R2 values above 0.90 and lower MAE and RMSE compared to CNN and GRU. The GRU model showed similar accuracy with faster training time, while CNN produced higher errors due to the dominant temporal nature of PV output. These results align with recent studies that emphasize selecting suitable deep learning architectures for time-series energy forecasting. This work highlights the benefit of integrating real local meteorological data with deep learning models in a scenario-based design and provides practical insights for regional grid operators and energy planners to reduce production uncertainty. Future studies can improve forecast reliability by testing hybrid models and implementing real-time adaptive training strategies to better handle extreme weather fluctuations.
Through a co-design and community-based participatory approach with Latino and NHPI communities, we identified key themes for an initial beta version of a mHealth app, including priorities around mental health, access to resources, and chronic disease prevention. Social and cultural connectedness and generational approaches emerged as important strategies for successful intervention design and long-term adoption. RR2-10.2196/preprints.73927.
Grid-stiffened design offers significant advantages in impact-resistant design for thin-walled structures. However, the inherent nonlinearities and high computational costs associated with impact analysis pose considerable challenges. This article introduces a novel impact-resistant optimization framework to address these challenges. First, a non-uniform grid-stiffened strategy is proposed, which can achieve the stiffener layout optimization for arbitrarily spatial structures. Then, the equivalent static loads method (ESLM) is used to transform dynamic loads into equivalent static loads (ESLs). A novel impact-resistant optimization framework is then constructed to enhance the impact resistance using gradient-based optimization algorithms by combining the non-uniform grid-stiffened strategy with ESLM. Finally, the effectiveness of the proposed method is validated using a flat plate model subjected to blast load and an aeroengine fan blade model subjected to bird strike. The optimization results demonstrate that the proposed method can improve the impact resistance while ensuring manufacturability.
The determination of unknown attacks remains a major challenge in WSN. Network Intrusion Detection (NIDS) is a proactive network security protection technology, which provides an effective defense system for WSN. NIDS heavily utilizes approaches for data extraction and Machine Learning (ML) to find anomalies. ML is an artificial intelligence subset that refers to a set of approaches allowing to learn from a preset dataset with improvement without human intervention. In terms of feature Importance. Particle Swarm Optimization (PSO) is a method used to select features in the dataset that contribute the most to predicting the target variable. Working with selected features instead of all the features reduces the risk of over-fitting, improves accuracy and decreases the training time. PSO technique selects optimal features from the preprocessed dataset. Residue Number System (RNS) is a numeral system representing integers by their values modulo several pairwise coprime integers called the moduli. This representation is allowed by Forward Conversion, which asserts that if N is the product of the moduli, there is, in an interval of length N, exactly one integer having any given set of modular values. The goal of the study is to provide NIDS with an attribute selection approach. PSO has been used for that purpose. This proposed feature selection method integrates RNS with the advantages of both empirical mode decomposition to retain most of the relevant features. The Network Intrusion Detection model PSO-RNS is being developed to identify any malicious activity in the network or any unusual behavior in the network, allowing the identification of the illegal activities. The proposed framework validated datasets, UNSW NB-15 to train the ensemble Machine Learning classifiers, KNN, Naïve Bayes and Logistic Regression as base classifiers while Random Forest as meta classifier, all been. stacked for feature selection with PSO optimization technique. In order to enhance the accuracy of the model, RNS is used to extract features from the dataset further using moduli set of {2n - 1, 2n, 2n +1}. The proposed PSO-RNS algorithm performs well in the benchmark function test and effectively guarantees the improvement of PSO feature selection approach. Our model achieved a reduced training time with the inclusion of RNS compared with PSO for (Naïve Bayes + KNN) + Random Forest: CASE A and KNN + Logistic Regression) + Random Forest: CASE B and improved accuracy. The experimental results show that the proposed intrusion detection model has good effects and practical application significance.
