Search

With the advancements in mobile technology, a large amount of data is generated by end devices, which has created a renewed interest in developing new AI-based applications for gaining insights into these data. However, most of these distributed applications need data aggregated at a central server, which has posed severe bandwidth, latency, security, and privacy issues. This paper presents OSGAN (One-Shot distributed learning algorithm using Generative Adversarial Networks), a generic framework that trains a generative adversarial network (GAN) at each client and uses GAN's generative capabilities to create sample data at the server. The server aggregates these data from various clients, builds a deep learning model and sends its parameters back to the clients in one communication round, i.e. the exchange of information between the clients and the server happens only once. In this paper, we present the design and implementation of OSGAN and evaluate its performance by comparing it with the state-of-the-art federated learning (FL) and central training algorithms for both IID and non-IID distribution of data. Our experiments on multiple datasets show that our proposed approach achieves a similar accuracy when compared with both FL and central training algorithms. Specifically, the accuracy drop with OSGAN is maintained within 2% for multiple datasets and multiple numbers of clients. Our results show that the proposed approach reduces the amount of data transfer by almost 98% when compared with federated learning and close to 80% when compared with the central learning approach thereby providing substantial benefit in terms of saving the bandwidth. [ABSTRACT FROM AUTHOR]

Big data and artificial intelligence (AI) technology are complicated systems that will continue developing in recent years. This paper implemented a data lake architecture to handle massive data and perform data analysis in a real-time system. Using a data lake and AI model, a NetFlow storage monitoring system was deployed to perform a platform that can cover the storage, query, analysis, and visualization of massive volumes of data. The big data platform was built on Cloudera, which utilized big data tools like Kafka, Spark, HBase, Hive, and Impala. In addition, we used Spark to develop network threat recognition models using distributed deep learning. Also, we used the deep neural network (DNN) to train the model. Then, we evaluated the model performance, which reached 94% accuracy while decreasing by 48% of training time. The results of the studies demonstrate that deep learning model training time is significantly shortened. Additionally, this system employs several configurations to assess the elements influencing accuracy and performance. The model is evaluated using the confusion matrix to demonstrate that it can accurately detect attack behavior in log data. Furthermore, we have developed a real-time log data monitoring and analysis system to demonstrate the proposed architecture. [ABSTRACT FROM AUTHOR]

Lung cancer is an uncontrolled growth of tissue causing a lump in the human lung. If lung cancer can be detected early, it can increase the survival rate. Therefore, a multi-classification approach of lung nodule detection with high computational effectiveness is required. In this paper, a multi-classification approach of lung nodule detection and classification is proposed using artificial intelligence on computed tomography (CT) scan images. Different preprocessing steps are applied for resizing, smoothing, and enhancement of the CT images. Then, two different approaches for feature extraction using VGG16 transfer learning and morphological segmentation are proposed. Morphological segmentation and feature extraction are applied for the segmentation of the region of interest and to extract the distinct features. Finally, the proposed deep learning architecture and seven different machine learning algorithms are applied on the preprocessed data and the extracted features for the classification of lung nodules into three classes: malignant, benign, and normal. It is observed that the stacked ensemble model of deep learning convolutional neural network (CNN) and VGG16 transfer learning models (CNN+VGG16) can achieve 99.55% accuracy using preprocessed data. It is also observed that all the ML algorithms perform with reasonably high accuracy using the low-dimensional morphological features. It is observed from the fivefold cross-validation results that logistic regression performs with 99.36% accuracy in 23.71 s time using the preprocessed data. Whereas, using the morphological features, k-nearest neighbor, and the support vector machine perform with the highest accuracy of 99.76% with very reduced computational time of 0.017 and 0.008 s, respectively. [ABSTRACT FROM AUTHOR]

As the world population continues to age, chronic diseases are on the rise. One of these diseases is stroke, which is a dangerous disease that can lead to many social and economic difficulties. Strokes can cause persistent neurological sequelae and physical disabilities; in some cases, motor function of the upper or lower body may be impaired, resulting in abnormal walking patterns or a loss of walking ability. These differences can be captured through walking patterns and gait information. There is therefore a need for research examining systems that can quickly detect signs of stroke based on human biosignals collected during daily life, such as motor function and the walking speed of the upper and lower limbs. By accurately predicting the early symptoms of stroke diseases, neural damage can be reduced with timely visits to medical institutions for treatment. In this paper, we designed and implemented a new AI-based system using real-time motion data for predicting stroke in the elderly. Initial data were collected from elderly Koreans while walking with attached wearable sensors. The sensors were placed on both shoulders and quadriceps. The data were processed, and we obtained a total of 12 motion attributes (angles and acceleration information). Predictive models using machine learning and deep learning algorithms were then constructed. The performance of the proposed system was verified with high prediction accuracies of 98.25% for the C4.5 decision tree model, 98.72% for RandomForest, 96.60% for XGBoost, and 98.99% for long short-term memory (LSTM). Hence, this paper provides a method for quickly and accurately predicting the early onset of stroke based on motion attribute values obtained while walking. [ABSTRACT FROM AUTHOR]

Microelectronics production failure analysis is an important step in improving product quality and development. In fact, the understanding of the failure mechanisms and therefore the implementation of corrective actions on the cause of the failure depend on the results of this analysis. These analyses are saved under textual features format. Then such data need first to be preprocessed and vectorized (converted to numeric). Second, to overcome the curse of dimensionality caused by the vectorisation process, a dimension reduction is applied. A two-stage variable selection and feature extraction is used to reduce the high dimensionality of a feature space. We are first interested in studying the potential of using an unsupervised variable selection technique, the genetic algorithm, to identify the variables that best demonstrate discrimination in the separation and compactness of groups of textual data. The genetic algorithm uses a combination of the K-means or Gaussian Mixture Model clustering and validity indices as a fitness function for optimization. Such a function improves both compactness and class separation. The second work looks into the feasibility of applying a feature extraction technique. The adopted methodology is a Deep learning algorithm based on variational autoencoder (VAE) for latent space disentanglement and Gaussian Mixture Model for clustering of the latent space for cluster identification. The last objective of this paper is to propose a new methodology based on the combination between variational autoencoder (VAE) for the latent space disentanglement, and genetic algorithm (GA) to find, in an unsupervised way, the latent variables allowing the best discrimination of clusters of failure analysis data. This methodology is called VAE-GA. Experiments on textual datasets of failure analysis demonstrate the effectiveness of the VAE-GA proposed method which allows better discrimination of textual classes compared to the use of GA or VAE separately or the combination of PCA with GA (PCA-GA) or a simple Auto-encoders with GA (AE-GA). [ABSTRACT FROM AUTHOR]

Semantic relatedness is one of the most significant aspects of natural language processing. It has been identified as a critical technology for developing intelligent systems like Siri, Microsoft Ice, Cortana, and Xiaoai. In 2014, SemEval ranked SR as the top task. While many existing studies have focused on analyzing the entailment of single phrases, advancements in deep learning have made it possible to analyze complete sentences or texts. While the natural parallelism of membrane computing has shown promise for data processing, harnessing this potential to advance semantic relatedness remains an open problem yet to be tackled. This paper proposes a novel Siamese Manhattan LSTM-SNP approach (SiMaLSTM-SNP) for the SR problem. The approach uses a collaborative Word2vec and 10-Layer Attention strategy to represent and extract sentence pairs and a Siamese LSTM-SNP structure to calculate the hidden states of sentences. The multi-head self-attention layer identifies text associations and redistributes hidden state weights. The last hidden state is extracted, and the relatedness score is calculated using the Manhattan distance. The experiments demonstrate that SiMaLSTM-SNP outperforms 17 classical SR baselines and 7 novel approaches on the standard datasets SICK and STS in terms of mean square error performance. This indicates that SiMaLSTM-SNP can accurately capture the semantic distinction between two sentences and effectively preserve their semantic information. [ABSTRACT FROM AUTHOR]

The production of traditional music is too complicated, consuming a lot of financial and human resources. Therefore, this paper aims to use artificial intelligence (AI) for songwriting and to explore the development and application of the Generative Adversarial Network (GAN) in smart music. An improved GAN-based Multi-Track Music (MTM)-GAN is established. The model is validated with the generation of 5 different music tracks for bass, drums, guitar, piano, and strings. The verification results are compared with the music generated by the existing Multi-Track Sequential GAN (MuseGAN) index evaluation method. The results show that many music clips generated by the MTM-GAN model are smooth and have a certain artistic aesthetic effect. Through the comparison of the two convergence curves of MuseGAN and MTM-GAN, when the penalty term is increased, the MTM-GAN of Consistency Term (CT) converges faster, and the training process is more stable. The numerical space of the parameter distribution obtained by the MTM-GAN-based music segment test is significantly smaller than that of MuseGAN. The probability of MTM-GAN overfitting is small. 62.8% of music listeners cannot distinguish the generated melody from the real melody. Therefore, the proposed model has the advantages of a more stable, more realistic, and faster fitting speed in music generation, indicating that the music generation method is effective. [ABSTRACT FROM AUTHOR]

With the Internet's rapid development and the increasing amount of netizens, social contradictions frequently manifest over the Internet. Public emergencies develop and spread constantly online. Thus, it is of great significance to reasonably address the Online Public Sentiment (OPS) in the current critical stage of social transformation. The aim is to create a safe and credible network environment and realize the modern transformation of the dynamic evolution of OPS in public emergencies. Firstly, this paper expounds on the blocking process of the OPS evolution on public emergencies according to the Internet of Things-native big data. Then, it discusses the algorithm process of the Long Short-Term Memory (LSTM) Neural Network (NN) model. Further, it optimizes the LSTM NN model using the Adaptive Momentum Estimation (Adam). Finally, it simulates and predicts the OPS evolution using Artificial Intelligence technology and big data. The results show that the Adam-optimized LSTM NN model can predict the hotness of OPS in the dynamic evolution with high prediction accuracy. In predicting OPS evolution, the Mean Relative Errors (MRE) of the proposed Adam-LSTM, LSTM, and Backpropagation NN models are 0.06, 0.10, and 0.14, respectively. The proposed Adam-LSTM model presents the least MRE on the hotness of OPS. The relevant governments can refer to model-predicted OPS evolution to control public emergencies and OPS through the IoT. Therefore, the proposed Adam-LSTM model is feasible for predicting the OPS hotness. The finding has particular research significance for employing the LSTM model under the IoT in predicting the OPS evolution in public emergencies. Lastly, the OPS on public emergencies can be effectively guided thanks to the proposed Adam-LSTM prediction model and time nodes. [ABSTRACT FROM AUTHOR]

There are many studies adopting artificial intelligence (AI) to develop core technologies for the future army but they are still at the level of basic research. It is expected that military power will be negatively affected by aging and declining population. In addition, as more than 500,000 agents will be dispatched to monitor combat scenes, the data sensed by each agent should be managed simultaneously recognize and evaluate the situation on the battlefield in real time. Despite increased complexity in the battlefield, current command system entirely rely on the experience and expertise of individual commanders, which severely restricts defense capabilities. Therefore, AI based military staff needs to be developed to identify potential threats that commanders are likely to miss, to develop smart command systems, and to provide data-driven rationale for commander's decisions. In this paper, we propose a deep AI military staff to support commander decision-making. Our proposed model is composed of four key parts: multi-agent based manned-unmanned collaboration architecture (MACA), robust tactical map fusion technology in poor environments (RTMF), hypergraph based representation learning (HRL) and space-time multi layer model for battlefields recognition (STBR). We design an architecture and generate dataset for training the core network. Simulation results are provided to demonstrate the performance of Deep AI military staff. [ABSTRACT FROM AUTHOR]

Deep learning compiler tool, Tensor Virtual Machine (TVM), has excellent deployment, compilation, and optimization capabilities supported by the industry following the vigorous growth in neural networks (NN). It has a unified intermediate representation (IR) format that can provide efficient compilation and portability. However, its high operational complexity requires considerable effort in development. For beginners with programming backgrounds, a new and easy-to-use design approach is needed. This paper proposes a visual concept approach that can execute artificial intelligence (AI) computing using block-based tools with AI knowledge. This research also develops a web-based NNBlocks framework that uses this approach to integrate with TVM. We conduct experiments to evaluate this approach: (1) interviewees assessed intuition through operating. (2) Interviewees answered a Usability Metric for User Experience (UMUX) to evaluate usability. (3) Interviewees answered the significance of the theme survey assessment. (4) The impact on the system was evaluated through experiments. The results indicate that interviewees respond positively to the intuitiveness of the framework. The usability evaluation of UMUX meets expectations. The theme survey shows that the framework is significant for AI learning. The experiments of the impact indicate that the framework will not burden the system. [ABSTRACT FROM AUTHOR]

Sparse linear algebra is central to many areas of engineering, science, and business. The community has done considerable work on proposing new methods for sparse matrix-vector multiplication (SpMV) computations and iterative sparse solvers on graphical processing units (GPUs). Due to vast variations in matrix features, no single method performs well across all sparse matrices. A few tools on automatic prediction of best-performing SpMV kernels have emerged recently and require many more efforts to fully utilize their potential. The utilization of a GPU by the existing SpMV kernels is far from its full capacity. Moreover, the development and performance analysis of SpMV techniques on GPUs have not been studied in sufficient depth. This paper proposes DIESEL, a deep learning-based tool that predicts and executes the best performing SpMV kernel for a given matrix using a feature set carefully devised by us through rigorous empirical and mathematical instruments. The dataset comprises 1056 matrices from 26 different real-life application domains including computational fluid dynamics, materials, electromagnetics, economics, and more. We propose a range of new metrics and methods for performance analysis, visualization, and comparison of SpMV tools. DIESEL provides better performance with its accuracy 88.2 % , workload accuracy 91.96 % , and average relative loss 4.4 % , compared to 85.9 % , 85.31 % , and 7.65 % by the next best performing artificial intelligence (AI)-based SpMV tool. The extensive results and analyses presented in this paper provide several key insights into the performance of the SpMV tools and how these relate to the matrix datasets and the performance metrics, allowing the community to further improve and compare basic and AI-based SpMV tools in the future. [ABSTRACT FROM AUTHOR]

In the past decade, social media networks have received much attention among ordinary people, agencies, and research scholars. Twitter is one of the fastest-growing social media tools. By means of the Twitter application on smartphones, users are able to immediately report events happening around them on a real-time basis. The information disseminated by millions of active users every day generates a new version of a dynamic database that contains information about various topics. Twitter data can be utilized as a major traffic data source along with conventional sensors. In this aspect, this paper presents a novel firefly algorithm-based feature selection with a deep learning model for traffic flow analysis (FFAFS-DLTFA) using Twitter data. The goal of FFAFS-DLTFA is to determine the class labels for tweets as relevant to traffic events. The proposed FFAFS-DLTFA encompasses several processes, such as preprocessing, feature extraction, feature selection, and classification. Primarily, tweets are preprocessed in several ways, such as tokenization, removal of stop words, and stemming. At the same time, three types of embedding vectors, unigram, bigram, and POS features, are used. In addition, the firefly algorithm (FFA) is applied for the optimal selection of feature subsets. Finally, a deep neural network (DNN) model is applied for the identification of tweets into three classes, namely, positive, neutral, and negative. The performance validation of FFAFS-DLTFA takes place using the benchmark Kaggle repository, and the results are inspected under different aspects. The experimental values demonstrate the better performance of FFAFS-DLTFA on the other techniques with the maximum accuracy of 98.83%. [ABSTRACT FROM AUTHOR]

The banking sector is on the eve of a serious transformation and the thrust behind it is artificial intelligence (AI). Novel AI applications have been already proposed to deal with challenges in the areas of credit scoring, risk assessment, client experience and portfolio management. One of the most critical challenges in the aforementioned sector is fraud detection upon streams of transactions. Recently, deep learning models have been introduced to deal with the specific problem in terms of detecting and forecasting possible fraudulent events. The aim is to estimate the unknown distribution of normal/fraudulent transactions and then to identify deviations that may indicate a potential fraud. In this paper, we elaborate on a novel multistage deep learning model that targets to efficiently manage the incoming streams of transactions and detect the fraudulent ones. We propose the use of two autoencoders to perform feature selection and learn the latent data space representation based on a nonlinear optimization model. On the delivered significant features, we subsequently apply a deep convolutional neural network to detect frauds, thus combining two different processing blocks. The adopted combination has the goal of detecting frauds over the exposed latent data representation and not over the initial data. [ABSTRACT FROM AUTHOR]

Since the advent of deep belief network deep learning technology in 2006, artificial intelligence technology has been utilized in various convergence areas, such as autonomous driving and medical care. Some services requiring fast decision making and action typically work seamlessly with edge computing service model. In autonomous driving of a connected vehicle with vehicle-to-everything (V2X) communication, roadside unit (RSU) acts as an edge computing device and it will make safer service by processing V2X messages fast, sent by vehicles or other devices. IEEE 1609.2 standard provides application message security technology to ensure the security and reliability of vehicle-to-vehicle communication messages. It uses elliptic curve digital signature algorithm (ECDSA) signatures based on the NIST p256 curve for message authenticity. In this paper, we investigate that RSU should be able to verify 3500 ECDSA signatures per second considering the expected maximum number of vehicles on nearby roads (e.g., during rush hour), message transmission rate, and IEEE 802.11p wireless channel capacity. RSU should satisfy this requirement without assistance of hardware-based cryptographic accelerator. For the requirement, we propose a hybrid approach of parallel ECDSA signature verification at high speed by using CPU and GPU, simultaneously. Moreover, we implemented the proposed method in various modern computing environments for RSU and edge computing devices. Through the experiments, we reach the conclusion that GPU can contribute to the required performance of ECDSA signature verification in RSU platform, which could not satisfy the above throughput only with CPU unit. The target platform with Intel Pentium E6500 CPU and GeForce GTX650 GPU can verify 5668 signatures per second with 30% utilization, while CPU in the platform can process only 2640 signatures. Even in a higher-performance edge computing device, we examine experimentally that the performance can be further improved by using the proposed hybrid approach. [ABSTRACT FROM AUTHOR]

Machine health monitoring systems are vital components of modern manufacturing industries. As advanced sensors collecting machine health-related data become commonplace, such systems have started adopting data-driven approaches to harness the collected data. However, dealing with noisy data and gleaning the spatial and temporal correlation within the data is a challenge. Extant literature focuses on the use of feature extraction to judge the state of normal and worn tools. However, the process of detecting the features of worn tools can often damage the workpiece material and result in machine downtime, which increases costs. Using recent developments in artificial intelligence, data features can be used to gauge the condition of tools in real-time. This paper proposes Holistic–Local Long Short-Term Memory (HLLSTM), a deep learning approach that adopts Long Short-Term Memory to predict tool wear based on holistic and local features. The data is divided into segments to learn short-term data features and further divided into holistic training and local training data to extract more implicit feature information to improve the accuracy of tool wear prediction. Experimental results indicate that HLLSTM can reduce the mean absolute error of real tool wear value two-fold and accurately predict tool wear. [ABSTRACT FROM AUTHOR]

At present, the security of the Internet of things (IoT) has aroused great concern in artificial intelligence area, Artificial Intelligence of Things (AIoT) are widely used in various intelligent surveillance scenarios. However, due to the weak interpretability of the model and high data security risks, developing a robust and explainable deep learning network framework for scene understanding under AIoT is extremely difficult. In addition, the fusion of IoT and AI also poses several challenges. To solve these difficulties, we develop a self-learning and explainable deep learning network toward the security of AIoT. The constructed system contains video collection, upload and display as well as data analysis and early warning operation at the embedded device end, and automatically recognizes the behaviors of scene by our developed visual recognition algorithms. In addition, the cloud computing platform can be controlled through our developed network. Our developed visual recognition algorithms contribute to three aspects. First, we propose a lightweight reinforcement learning network model by extracting spatial–temporal feature of different behavior characteristic. Then, we propose a self-paced learning framework through fusing the deep reinforcement learning and transfer learning. Finally, we propose a multi-perspective deep transfer learning model to solve the problem of weak explanation of model. The experimental results show that our proposed model is able to provide high interpretability of model and outperforms the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The Internet of Things (IoT) is driving the digital revolution. AlSome palliative measures aremost all economic sectors are becoming "Smart" thanks to the analysis of data generated by IoT. This analysis is carried out by advance artificial intelligence (AI) techniques that provide insights never before imagined. The combination of both IoT and AI is giving rise to an emerging trend, called AIoT, which is opening up new paths to bring digitization into the new era. However, there is still a big gap between AI and IoT, which is basically in the computational power required by the former and the lack of computational resources offered by the latter. This is particularly true in rural IoT environments where the lack of connectivity (or low-bandwidth connections) and power supply forces the search for "efficient" alternatives to provide computational resources to IoT infrastructures without increasing power consumption. In this paper, we explore edge computing as a solution for bridging the gaps between AI and IoT in rural environment. We evaluate the training and inference stages of a deep-learning-based precision agriculture application for frost prediction in modern Nvidia Jetson AGX Xavier in terms of performance and power consumption. Our experimental results reveal that cloud approaches are still a long way off in terms of performance, but the inclusion of GPUs in edge devices offers new opportunities for those scenarios where connectivity is still a challenge. [ABSTRACT FROM AUTHOR]

Nowadays, deep learning is one of the popular research areas of the computer sciences, and many deep networks have been proposed to solve artificial intelligence and machine learning problems. Residual networks (ResNet) for instance ResNet18, ResNet50 and ResNet101 are widely used deep network in the literature. In this paper, a novel ResNet-based signal recognition method is presented. In this study, ResNet18, ResNet50 and ResNet101 are utilized as feature extractor and each network extracts 1000 features. The extracted features are concatenated, and 3000 features are obtained. In the feature selection phase, 1000 most discriminative features are selected using ReliefF, and these selected features are used as input for the third-degree polynomial (cubic) activation-based support vector machine. The proposed method achieved 99.96% and 99.61% classification accuracy rates for gender and activity recognitions, respectively. These results clearly demonstrate that the proposed pre-trained ensemble ResNet-based method achieved high success rate for sensors signals. [ABSTRACT FROM AUTHOR]

Predictive maintenance is an invaluable tool to preserve the health of mission critical assets while minimizing the operational costs of scheduled intervention. Artificial intelligence techniques have been shown to be effective at treating large volumes of data, such as the ones collected by the sensors typically present in equipment. In this work, we aim to identify and summarize existing publications in the field of predictive maintenance that explore machine learning and deep learning algorithms to improve the performance of failure classification and detection. We show a significant upward trend in the use of deep learning methods of sensor data collected by mission critical assets for early failure detection to assist predictive maintenance schedules. We also identify aspects that require further investigation in future works, regarding exploration of life support systems for supercomputing assets and standardization of performance metrics. [ABSTRACT FROM AUTHOR]

Today, the internet of things (IoT) is becoming more common and finds applications in several domains, especially in the healthcare sector. Due to the rising demands of IoT, a massive quantity of sensing data gets generated from diverse sensing devices. Artificial intelligence (AI) techniques are vital for providing a scalable and precise analysis of data in real time. But the design and development of a useful big data analysis technique face a few challenges, like centralized architecture, security, and privacy, resource constraints, and the lack of adequate training data. On the other hand, the rising blockchain technology offers a decentralized architecture. It enables secure sharing of data and resources to the different nodes of the IoT network and is promoted for removing centralized control and resolving the problems of AI. This study develops an optimal deep-learning-based secure blockchain (ODLSB) enabled intelligent IoT and healthcare diagnosis model. The proposed model involves three major processes: secure transaction, hash value encryption, and medical diagnosis. The ODLSB technique comprises the orthogonal particle swarm optimization (OPSO) algorithm for the secret sharing of medical images. In addition, the hash value encryption process takes place using neighborhood indexing sequence (NIS) algorithm. At last, the optimal deep neural network (ODNN) is applied as a classification model to diagnose the diseases. The utilization of OPSO algorithm for secret sharing and optimal parameter tuning process shows the novelty of the work. We carried out detailed experiments to validate the outcome of the proposed method, and several aspects of the results are considered. At the time of the diagnosis process, the OPSO-DNN model has yielded superior results, with the highest sensitivity (92.75%), specificity (91.42%), and accuracy (93.68%). [ABSTRACT FROM AUTHOR]

Given the growth and availability of computing power, artificial intelligence techniques have been applied to industrial equipment and computing devices in order to identify abnormalities in operation and predict the remaining useful life (RUL) of equipment with superior performance than traditional predictive maintenance. In this sense, this research aims to develop a neural network applied to predictive maintenance in mission critical supercomputing environments (MCSE) using deep learning techniques to predict the RUL of an equipment before the occurrence of failures, by using real historical unlabeled data, which were collected by sensors installed in a supercomputing environment. The method was developed using a hybrid approach based on a combination of Fully Convolutional Neural Network, Long Short-Term Memory and Multilayer Perceptron. The results presented a Pearson R of 0.87, R 2 of 0.77, Factor of 2 of 0.89, and Normalized mean square error of 0.79, considering the predicted RUL value and the observed RUL value for the pre-failure behavior moments of the equipment. Thus, we can conclude that the developed approach had good performance to predict the RUL, increasing the ability to anticipate the failure situation of the MCSE, further increasing its availability and operating time. [ABSTRACT FROM AUTHOR]

With the advancement of artificial intelligence (AI) and machine learning techniques, the diseases can be early diagnosed and detected with the help of known healthcare data of similar diseases. However, the analysis of healthcare data becomes challenging due to the imbalance behavior of such data. Such behavior has a skewed nature and the samples of imbalanced dataset contain the majority and minority instances. The quantity of majority instances is much higher than that of the minority instances. Thus, the classifier always gets biased toward the majority instances while classifying an unknown instance. This is not a perfect deal by any of the classifiers especially for disease diagnosis. Therefore, it is important to balance the dataset by considering both the majority and minority class samples. In this context, Genetic Algorithm based on Autoencoder (GAAE) model is proposed to process the imbalance data. Initially, an autoencoder is trained by the help of genetic operators and both the majority and minority samples. The chromosome is efficiently designed to represent an autoencoder. The error function is designed as the fitness function with the help of ensemble of classifiers. The optimized autoencoder generates the synthetic data of minority class to balance the data. Once the imbalanced data is balanced using the proposed GAAE, then the feature selection is done based on the correlation coefficient. Then various classifiers, multi-layer perceptron (MLP), k-nearest neighbor (k-NN), C4.5 decision tree (DT) and random forest (RF) are employed for the data classification. An extensive simulation is performed on the proposed approach and comparison is done with the existing approaches. It is observed that consideration of both the majority and minority samples to generate the synthetic data helps the classifiers to perform better. In addition, statistical analysis has also been performed. [ABSTRACT FROM AUTHOR]

With the continuous upgrading and improvement in the Internet and terminal equipment, many instant music videos share information with users through social platforms. This study explores the impact of new media technology on the content of instant music videos on the Internet under Artificial Intelligence (AI) technology to effectively distinguish the elegant and vulgar short videos and improve the quality of short videos on the Internet. Obscene and harmful instant music videos in the massive data are the bottleneck for its development. An improved deep learning model is proposed based on OPEN_NSFW using the AI image detection system technology of the Internet of Things with a powerful processing ability to image information. Experiments demonstrate that this model significantly reduces the false positive rate and improves the recall compared with the traditional machine learning computing model. Besides, it improves the accuracy when discriminating whether the publisher's head image involves eroticism. In addition, this model can identify and classify the main content of instant music videos to optimize the content. This work provides the characteristic basis for the algorithm to judge and protect the original content. Combining algorithm recommendations and strengthening manual intervention promotes online instant music videos' sustainable and healthy development. These findings can provide an excellent technical guarantee and experimental references for the standardized development of the instant music video industry in the future. [ABSTRACT FROM AUTHOR]

Population size has made disease monitoring a major concern in the healthcare system, due to which auto-detection has become a top priority. Intelligent disease detection frameworks enable doctors to recognize illnesses, provide stable and accurate results, and lower mortality rates. An acute and severe disease known as Coronavirus (COVID19) has suddenly become a global health crisis. The fastest way to avoid the spreading of Covid19 is to implement an automated detection approach. In this study, an explainable COVID19 detection in CT scan and chest X-ray is established using a combination of deep learning and machine learning classification algorithms. A Convolutional Neural Network (CNN) collects deep features from collected images, and these features are then fed into a machine learning ensemble for COVID19 assessment. To identify COVID19 disease from images, an ensemble model is developed which includes, Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), Decision Tree (DT), Logistic Regression (LR), K-Nearest Neighbor (KNN), and Random Forest (RF). The overall performance of the proposed method is interpreted using Gradient-weighted Class Activation Mapping (Grad-CAM), and t-distributed Stochastic Neighbor Embedding (t-SNE). The proposed method is evaluated using two datasets containing 1,646 and 2,481 CT scan images gathered from COVID19 patients, respectively. Various performance comparisons with state-of-the-art approaches were also shown. The proposed approach beats existing models, with scores of 98.5% accuracy, 99% precision, and 99% recall, respectively. Further, the t-SNE and explainable Artificial Intelligence (AI) experiments are conducted to validate the proposed approach. [ABSTRACT FROM AUTHOR]

The National Basketball Association's (NBA) tournaments are watched by people all over the world via the Internet of Things (IoT). Players who score more points are regarded as star players and are part of the focus of competition for each team. The purpose of this research is to predict the score of NBA league players based on player data using deep learning methods. Various algorithms were compared, the best model for score predictions was obtained, and a comparative analysis of the team was provided to the players. The prediction effect and average MAPE index of limit gradient boosting (XGB) are better than that of other methods. It can be seen that the XGB method has a better fit to the regression problem and high interpretability. Thus, it has the ability to accurately predict and evaluate NBA score predictions. The results can support the team's decision-making in player management. [ABSTRACT FROM AUTHOR]

The present work aims to construct knowledge systems of different specialties, provide personalized learning approaches for students, improve the teaching effectiveness of teachers and learning interests of students, and help students accurately diagnose their learning problems. First, the existing issues of students' chemistry learning and knowledge systems are analyzed. Then, a personalized learning model is constructed based on the distributed computing method of the Internet of Things (IoT) and the clustering algorithm of deep learning (DL). This model realizes the effective classification of students by the DL algorithm according to the students' learning data and designs diverse customized teaching contents in line with the performance of IoT. Finally, this model's effectiveness is verified through the data analysis of 2019 chemistry students at M University. The research results indicate that 96.67% of students express satisfaction with the learning effect of the personalized learning model, and 100% of learners are satisfied with the various forms of teaching resources offered by this model. Besides, this model's accuracy can reach 85% on the personalized learning platform based on IoT and DL algorithms. Compared with the latest research model, this model has a better performance in achievement prediction and customized recommendation. The test results of the actual effect demonstrate that the personalized learning system has achieved expected outcomes, significantly enhancing students' understanding of knowledge with medium difficulty. In addition, learners are delighted with the model. They believe that the learning resources in the model can meet their learning needs, and they are willing to recommend the course to other learners actively. This model is of significant practical value to promote the development of IoT and DL technology in professional learning. This exploration innovatively stratifies learners' level of understanding and provides personalized learning resources meeting the current cognitive competence of students to enhance learners' knowledge and achieve personalized learning. [ABSTRACT FROM AUTHOR]

The purpose is to study the applicability of digital and intelligent real-time Image Processing (IP) in fitness motion detection under the environment of the Internet of Things (IoT). Given the absence of real-time training standards and possible workout injury problems during fitness activities, an intelligent fitness real-time IP system based on Deep Learning (DL) is implemented. Specifically, the keyframes of the real-time images are collected from the fitness monitoring video, and the DL algorithm is introduced to analyze the fitness motions. Afterward, the performance of the proposed system is evaluated through simulation. Subsequently, the Noise Reduction (NR) performance of the proposed algorithm is evaluated from the Peak Signal-to-Noise Ratio (PSNR), which remains above 20 dB for seriously noisy images (with a noise density reaching up to 90%). By comparison, the PSNR of the Standard Median Filter (SMF) and Ranked-order Based Adaptive Median Filter (RAMF) algorithms are not higher than 10 dB. Meanwhile, the proposed algorithm outperforms other DL algorithms by over 2.24% with a detection accuracy of 97.80%; the proposed system can adaptively detect the fitness motion, with a transmission delay no larger than 1 s given a maximum of 750 keyframes. Therefore, the proposed DL-based intelligent fitness real-time IP algorithm has strong robustness, high detection accuracy, and excellent real-time image diagnosis and processing effect, thus providing an experimental reference for sports digitalization and intellectualization. [ABSTRACT FROM AUTHOR]

COVID-19 has affected every individual physically or physiologically, leading to substantial impacts on how they perceive and respond to the pandemic's danger. Due to the lack of vaccines or effective medicines to cure the infection, an urgent control measure is required to prevent the continued spread of COVID-19. This can be achieved using advanced computing, such as artificial intelligence (AI), machine learning (ML), deep learning (DL), cloud computing, and edge computing. To control the exponential spread of the novel virus, it is crucial for countries to contain and mitigate interventions. To prevent exponential growth, several control measures have been applied in the Kingdom of Saudi Arabia to mitigate the COVID-19 epidemic. As the pandemic has been spreading globally for more than a year, an ample amount of data is available for researchers to predict and forecast the effect of the pandemic in the near future. This article interprets the effects of COVID-19 using the Susceptible-Infected-Recovered (SIR-F) while F-stands for 'Fatal with confirmation,' age-structured SEIR (Susceptible Exposed Infectious Removed) and machine learning for smart health care and the well-being of citizens of Saudi Arabia. Additionally, it examines the different control measure scenarios produced by the modified SEIR model. The evolution of the simulation results shows that the interventions are vital to flatten the virus spread curve, which can delay the peak and decrease the fatality rate. [ABSTRACT FROM AUTHOR]

With the recent emergence of artificial intelligence (AI) technology, autonomous vehicle industry has rapidly adopted this technology to investigate self-driving systems based on AI technology. Although autonomous driving is frequently used in high-speed environments, most studies are conducted on low-speed driving on complex urban roads. Currently, most commercialized self-driving cars in SAE autonomous driving level 2 provide practical performance on high-speed roads using various sensors. However, these systems have to process huge sensor data and apply complex control algorithms. Recently, studies have been conducted on the use of image-based end-to-end deep learning to control autonomous driving systems that can be configured at a low cost without expensive sensors and complex processes. In this study, we proposed an autonomous driving control system using a novel end-to-end deep learning model for high-speed environments, and also compared the performance of the proposed system with NVIDIA end-to-end driving system. [ABSTRACT FROM AUTHOR]

In recent years, object detection technology with artificial intelligence has been applied in many fields. This study uses a deep learning method to train an identification model to classify and browse pictures of the 600 different kinds of birds in Taiwan. To enhance the accuracy of identification and classification of these birds, we propose an automatic extraction system that can obtain training data by visiting public social media pages. We also develop mobile apps that allow users to take pictures of birds and upload them to an identification server to enable real-time identification and provide training data. These mobile apps are sent candidate bird pictures by the identification server to allow users to confirm and give feedback when the confidence level of identification is within a critical range. The bird pictures are then used as training data, and the identification model is periodically retrained to optimise the model. We also use natural language processing technology to enhance the level of confidence in image identification. The features of the birds' appearance are described in words and candidate birds are obtained through image identification and used to readjust the adopted weight values. The proposed identification system gives a relatively high identification accuracy due to the use of deep learning object detection. [ABSTRACT FROM AUTHOR]