Optical Memory and Neural Networks最新文献

Twitter has millions of active users and is a significant microblogging platform. These users use Twitter to give their thoughts on various events using hashtags also to make status updates known as tweets. As a result, Twitter is regarded as a significant real-time streaming source as well as a reliable and accurate opinion indicator. Due to Twitter’s massive data generation volume, it is challenging to manually scan the entire collection. Given the massive volume of data supplied by Twitter, it is challenging to manually scan the entire collection. So, a hybrid deep learning algorithm is developed to analyse the sentiment of the user. This research incorporates a variety of techniques like pre-processed using tokenization, stop word removal, stemming, Removal of hyperlinks and numbers, Abbreviation extending and spell correction. After that, use Semantic Lexicons with Puffer Fish Optimized GLOVE (SLPFOG) to extract features and convert words into vectors. The, reduce the dimension of the extracted features by applying the Laplacian Eigen map. To forecast the user sentiment of Twitter Big data, a hybrid Hopfield Neural Network—Bidirectional Gated Recurrent Unit (HNN-BiGRU) technique was created. The proposed hybrid HNN-BiGRU approach has an accuracy of 96%, specificity of 99%, NPV of 99% and MCC of 97%. Thus, the hybrid deep learning algorithm is the best option for sentimental analysis of twitter big data because they achieves relatively high accuracy with respect to basic algorithms without sacrificing the interpretability of the learning results.

The default mode network (DMN), also referred to as the “Passive Mode Brain Network” (PMBN), serves as a network of active brain regions while restfully stated. An abnormal homogeneity of the DMN network has been implicated in the first episode of schizophrenia, a mental disorder characterized by perceptual disturbances. This study aimed to investigate the activity and functional connectivity of the DMN in female schizophrenia patients using an innovative brain activity localization technique known as the “Virtually implanted electrode”. EEG was registered in 22 female patients diagnosed with schizophrenia, including 17 cases of F20, 3 cases of F23, and 22 healthy controls, being in a state of quiet wakefulness. The results indicated a complex system of changes in schizophrenia patients compared to controls, attributed to weakening connections originating from structures with reduced activity and reinforcing of other connections, including inhibitory ones. These findings underscore the neurobiological basis of schizophrenia, investigating the DMN.

In this work, we analyze the topological charge (TC) of finite superposition of optical vortices (OVs) with a Gaussian envelope. In the source plane, the superposition under study is theoretically and numerically shown to have the TC equal to the number of zeros of a complex polynomial of degree n, where n is the largest TC of the constituent OVs found inside and on a unit-radius circle. Meanwhile upon free space propagation, the TC of the superposition always equals n. We reveal that if, in absolute values, the coefficient of a superposition term with TC = k is larger than the sum of all the rest superposition coefficients, then k zeros occur inside the unit-radius circle, with the total TC of the superposition being equal to k (k ≤ n) in the source plane. If all the coefficients are equal to each other in the absolute value, then, in the source plane, TC takes a value of n/2, before returning to the value of n upon propagation. In this case, extra zeros of the superposition of OVs occur almost at once, at a subwavelength distance from the source plane, with the distance from the optical axis being larger than the radius of an aperture limiting the source field.

Vehicular Ad Hoc Network (VANET) has become a revolutionary and creative technology that serves as an essential part of Intelligent Transportation Systems (ITS). However, due to their wireless nature and complex operating environment, VANETs are vulnerable to a range of malicious user assaults. It is critical to identify intrusions in the VANET system in order to provide reliable and secure communication among all of the system’s vehicles. Traditional methods are no longer effective due to some limitations like lack of data, interpretability and imbalance classes. Therefore, the proposed approach developed an enhanced deep RL routing (EDRL) with SA-BiLSTM for the detection of intrusion and created a secure VANET system employing modular Homomorphic encryption. In this proposed model, consider if any incident happens on the road, vehicles in that sector are grouped by utilizing the Improved K harmonic means clustering algorithm (IKHM), and the CH is determined according to its minimal distance and highest energy using the Greater Cane Rat Algorithm (GCRA) optimization. The EDRL routing technique is then used to exchange the data to RSU for choosing the appropriate route. RSU discovered the different types of attack and non-attack using Self Attention-Based Bidirectional Long Short-Term Memory (SA-BiLSTM) classifier. Then the non-attack data are encoded using the Modular Homomorphic Encryption (ModHE) and uploaded in the cloud to intimate the warning message to the vehicular networks. The proposed model’s performance parameters are examined, and the results show that, for 500 vehicle nodes, the outcomes are 82.2% PDR, 13.65J energy usage, 20.3% routing overhead, 18.7 mbps throughput, and 11.22 delay. Accuracy, hit rate, and PPV are assessed at 96.3, 96.7, and 95.8%, respectively, for attack detection. Furthermore, the execution time and encryption take 16.63 and 46.03 milliseconds, respectively. The mentioned results demonstrated that the proposed framework outperformed earlier methods in providing a remarkably energy-efficient as well as secure V2X communication network.

Besides scalar optical vortices that have a topological charge (TC), helical wave front, and carry an orbital angular momentum (OAM) that can be transferred to particles and rotate them along circular trajectories, polarization optical vortices are also known, whose polarization state in the beam section changes with the azimuthal angle. Such vortices are polarization singularities that are described by indices, similar to the TC. However, polarization OAM for polarization vortices still has not been considered, although laser beams with inhomogeneous polarization can perform spiral mass transport in polarization-sensitive media. In this work, we consider two possible definitions of the polarization OAM. One OAM is proportional to the azimuthal velocity of changing direction of linear polarization vector, whereas the other (hybrid OAM) is proportional to the azimuthal velocity of changing ellipticity degree of the polarization ellipse. For instance, the normalized polarization OAM is equal to the order of a cylindrical vector beam and also equals the order of Poincaré beam.

COVID-19 was thought to be the most lethal and devastating disease for humans caused by the novel coronavirus currently. Accurate diagnosis may lead to earlier COVID-19 discovery and lower patient mortality, especially in instances without evident symptoms. The majority of the time, chest X-ray (CXR) images are used to diagnose this illness. Patients who are infected with coronavirus exhibit symptoms that were very similar to those of pneumonia, and the virus targets body’s respiratory organs, making breathing difficult. This paper presented a hybrid VGG19-SVM model for identifying COVID-19 patients in CXR based on wild horse optimizer (WHO) based K-means segmentation to address these problems. The proposed segmentation algorithm comprises four phases such as data gathering, pre-processing, segmentation and COVID-19 detection. CXR data were gathered from medical Internet of Things (IoT) devices. Image pre-processing was performed with the assistance of image resizing, Markov random field (MRF) and adaptive gamma correction (AGC). Then, the proposed WHO based K-clustering is used to segment the affected portion of lung CXR effectively. The hybrid classification approach is introduced based on the combination of VGG19 and SVM, which is employed to classify if the patient is in normal condition either COVID-19, pneumonia or tuberculosis. Thus, various existing methods such as VGG19, AlexNet, VGG16 and GoogleNet are taken in this analysis. The proposed VGG19-SVM attained 0.96 of F1_score, 0.97 of NPV, 0.07 FNR and 0.008 of FPR, when compared to the existing methods obtained better findings using DL techniques. This shows the effectiveness of the proposed WHO based K-means clustering algorithm and hybrid VGG19-SVM model which can be useful for segment the CXR images.

Electrocardiographic (ECG) signals are vital for identifying and assessing cardiac problems. However, a variety of noises can contaminate ECG data, which affects the utility of ECG signals in application. Errors may be induced by patient movements, electromagnetic noise in surrounding devices, or muscle contraction artifacts. Traditional methods have often struggled with balancing effective noise reduction while preserving critical signal details, leading to compromised diagnostic accuracy. Various methods like adaptive filtering, wavelet methods, and EMD are used to denoise ECG signals to prevent noisy inference, but they may suffer with non-stationary noise or complex interference patterns. To address the aforementioned difficulties, an optimized deep learning approach and smoothing filter is designed for effectively increase the quality and reduce noise in the ECG signal. Initially, noisy ECG signals are obtained from the ECG heartbeat categorization dataset. The collected ECG raw signal is decomposed by the Multivariate dynamic mode decomposition (MDMD) technique for obtaining both high-frequency and low-frequency components of multivariate time-series data. Then, noise existing in both high frequency components is effectively removed by applying the LU-Net technique. Manta ray foreign optimization (MRFO) approach is utilized to select the learning rate and batch size of the LU-Net classifier in an optimal manner. The Integrate-and-Fire Time Encoding Machine (IF-TEM) method is used to reconstruct the denoised ECG signal. Signal sparsity assisted signal smoothing (SASS) approach is used to denoise and enhance the quality of ECG signal. The proposed MDLUTESS denoising method is compared with existing methods and its effectiveness is assessed using performance metrices like SNR, PSNR, MSE were 42, 53 dB, and 0.0017. Thus the proposed method successfully eliminates noise from the ECG signals.

In this study, the results of solving a “synthesis–properties” type problem using artificial neural networks have been presented. The purpose of the study has been to determine the optimal conditions for synthesis of carbon dots to obtain nanoparticles with a given luminescence quantum yield (QY). Carbon dots were synthesized by hydrothermal synthesis from citric acid and ethylenediamine at various conditions. A multilayer perceptron (MLP) type artificial neural network was used to approximate the dependence of the target variable (luminescence QY) on the synthesis parameters. The neural network approach was successfully applied to the spectral data of a set of carbon dots of 343 samples to determine the optimal conditions for their hydrothermal synthesis from citric acid and ethylenediamine while varying the precursor ratio, temperature and reaction time over wide ranges to obtain nanoparticles with a given luminescence QY. Optimal carbon dots synthesis parameters to maximize the luminescence QY at 350 nm have been determined. Testing of the proposed neural network approach on an independent database of spectral data specially synthesized for this purpose showed good agreement between the results obtained using MLP and the experimentally measured values of the QY (the root-mean-squared error of the QY prediction was 2.14%).

The red palm weevil (RPW) represents a significant danger to palm trees farms all over the world, which will result in considerable financial losses. The absence of apparent signs until the death of the palm tree makes it difficult to identify RPW infections at an early stage. The prompt detection of RPW diseases is further complicated by large-scale farms. In order to accomplish early detection of RPW using image analysis, this paper proposed a RPW classification model based on the proposed modified ResNet-34 deep learning architecture. A dataset of 483 images is used to assess the model’s performance. For the assessment, two different dataset settings are used. In the initial dataset setup, images are divided into three groups: adults, eggs, and Pupae. Four additional categories are added to the classification in the second dataset setup: female adults, male adults, eggs, and pupae. Experimental findings show the usefulness of the proposed model, with a remarkable total accuracy of 98% for both dataset setups. These results highlight the value of using the modified ResNet-34 architecture for the early detection of RPW. Moreover, the findings demonstrated that the proposed model offers great potential for decreasing the negative effects of RPW on palm tree farms and preventing financial losses in the agriculture sector.

Scrutiny of consumer tweets posted on social media is found to be indispensable for numerous business applications. In this manner, the model of big data analytics is applied in processing data and analyzes it to predict consumer behavioral patterns on social media. Different machine learning algorithms have gathered consumer data to analysis consumer behavior. Conventional methods are unable to discover extreme hidden patterns and require to be enhanced to produce more accurate behavioral patterns. In this work a hybrid method called, proposed Bouldin Correlation Clustering and Gradient Extreme Learning Machine (BCC-GELM) method to perform the consumer behavior analysis in social network with big data. The BCC-GELM method in hybrid model split into two modules. At first, Davis-Bouldin Index-based Correlation Clustering selects clusters with most edges within clusters as positive (i.e., similar information) while most edges between clusters as negative (i.e., dissimilar information), therefore minimizing the error rate. Consumer previous behavioral characteristics and twitter messages are analyzed by means of focal points (i.e., cluster center) via Davis-Bouldin Index. Subsequently, Stochastic Gradient Descent Extreme Learning Machine yields good results by considering distribution of tweets, therefore paving way for predicting consumer behavioral patterns in an optimal manner. The performance of BCC-GELM method is evaluated using experimental analysis and comparison is also made with traditional consumer behavioral pattern methods. The findings demonstrate that BCC-GELM method performs well than the traditional consumer behavioral pattern methods in terms of 9% of clustering accuracy, 45 and 54% of clustering time using without and with preprocessing (percent), 23% of clustering overhead and 46% of error rate.