Face annotation plays a crucial role in the field of computer vision. Its purpose is to accurately label the faces that appear in an image. The effectiveness of face annotation relies heavily on the representation of facial features, such as color, texture, and shape. Deep texture features, in particular, play a significant role in face annotation systems. It is worth noting that different individuals can possess similar texture features, which can impact the performance of annotation. Therefore, this study addresses the enduring complexity of face similarity by introducing an innovative approach called the Deep Learning-based Texture Feature (DLTF) through the utilization of the efficient deep learning model known as the Residual Network (ResNet). Despite the variations in poses, lighting, expressions, and occlusions that can greatly alter faces, ResNet’s deep architecture and feature retention capabilities make it resilient to these changes, ensuring consistent and accurate annotations under diverse conditions. Experimental results obtained from the IMFDB, LFW, and Yahoo datasets demonstrate that the proposed DLTF is the most effective description of deep texture features, leading to improved face naming performance. Furthermore, the proposed DLTF enhances the efficiency of the face-naming task by effectively addressing real-life challenges.

This research intends to create a novel approach for solving fractional differential equations (FDEs) of both linear and nonlinear types utilizing the fractional shifted Legendre neural network alongside the extreme machine learning (ML) algorithm. The neural network serves as a trial solution to formulate the loss function where it is used to train the neural network through an extreme learning machine (ELM) to obtain the solution. The new approach is applied to the linear differential equations and the unknown coefficient matrix is identified using the psuedoinverse of the activation function output matrix whereas the nonlinear equations are handled by employing nonlinear least square perturb algorithm. An algorithm is provided for solving the FDEs. Theoretical analysis such as convergence, stability, and error estimates are studied. The proposed method is tested on various problems and is compared with the other algorithms discussed in the literature.

With the rapid proliferation of machine learning applications in cloud computing environments, addressing crucial challenges concerning energy efficiency becomes pressing, including addressing the high power consumption of such workloads. In this regard, this work focuses much on the development of an energy-aware scheduling and task assignment algorithm that, while optimizing energy consumption, maintains required performance standards in deploying machine-learning applications in cloud environments. It therefore, pivots on leveraging online reinforcement learning to deduce an optimal planning and allocation strategy. This proposed algorithm leverages the capability of RL in making sequential decisions with the aim of achieving maximum cumulative rewards. The algorithm design and its implementation are examined in detail, considering the nature of workloads and how the computational resources are utilized. The algorithm’s performance is analyzed by looking into different performance metrics that assess the success of the model. All the results indicate that energy-aware scheduling combined with task assignment algorithms are bound to reduce energy consumption by a great margin while meeting the required performance for large-scale workloads. These results hold much promise for the improvement of sustainable cloud computing infrastructures and consequently, to energy-efficient machine learning. The future research directions involve enhancing the proposed algorithm’s generalization capabilities and addressing challenges related to scalability and convergence.

The increasing incidence of mental health issues demands innovative diagnostic methods, especially within digital communication. Traditional assessments are challenged by the sheer volume of data and the nuanced language found on social media and other text-based platforms. This study seeks to apply machine learning (ML) to interpret these digital narratives and identify patterns that signal mental health conditions. We apply natural language processing (NLP) techniques to analyze sentiments and emotional cues across datasets from social media and other text-based communication. Using ML, deep learning, and transfer learning models such as bidirectional encoder representations (BERTs), robustly optimized BERT approach (RoBERTa), distilled BERT (DistilBERT), and generalized autoregressive pretraining for language understanding (XLNet), we assess their ability to detect early signs of mental health concerns. The results show that BERT, RoBERTa, and XLNet consistently achieve over 95% accuracy, highlighting their strong contextual understanding and effectiveness in this application. The significance of this research lies in its potential to revolutionize mental health diagnostics by providing a scalable, data-driven approach to early detection. By harnessing the power of advanced NLP models, this study offers a pathway to more timely and accurate identification of individuals in need of mental health support, thereby contributing to better outcomes in public health.

Globally, one of the most fundamental applications of time-series data is stock market forecasting, where each and every second is crucial and analysis is unpredictable, posing a significant challenge towards predicting its trend. Various research is underway for the analysis of the stock market using various time series algorithms such as long-short-term memory (LSTM), recurrent neural network (RNN). Here, in this article, we have proposed a novel approach of multiple layers including liquid time constant, the linear first order dynamics with nonlinear interlinked gates which improves performance of time series; after which we included liquid neural network and forecasted the results. Then, we enhanced the algorithm using Boruta feature selection, where we have selected only the required columns and ranked them as per requirement. Furthermore, for more accurate results, we have selected three different sets of historical data for five large cap sectors (Alphabet, Apple, Blackrock, JP Morgan, and IBM) after which we observed the R-square values along with all the performance evaluation and error evaluation results for all the individual dataset in two phases, i.e., before and after using the Boruta feature selection method for 5, 10, and 15 years data, we observed the best results for LSTM (0.9443) and RNN (0.9752). Liquid neural networks, being the most efficient and accurate, gave the best R-square value of 0.9835.

Automatic text summarization (ATS) deals with compressing a long document into a shorter version, retaining the key ideas of the original document. It aims to tackle the problem of information overload resulting from the continuous creation of documents. ATS helps people make news feeds, meeting reports, summarized legal and financial documents, study tools, article outlines, social media analysis, and so on. Extractive text summarization (ETS) is a type of ATS that creates summaries (extracts) by choosing sentences from the source without modifying the semantics and structure of the sentences. The main problem in ETS is to select the appropriate sentences that will constitute a coherent, concise, and nonredundant summary that covers the entire document. The literature suggests many optimization-based algorithms for the selection problem. Some authors have used binary particle swarm optimization (BPSO), genetic algorithm (GA), and artificial bee colony (ABC) algorithm to address the selection problem. However, these algorithms suffer from the problem of premature or slow convergence. This article proposes an ETS algorithm using BPSO and masked GA (EBPGA). The novelty of the proposed system is the usage of the masked GA (MGA), which leverages the properties of the document, namely, coherence, redundancy, relevance, and coverage factors, along with domain-specific mutation and crossover to tackle the premature convergence of the BPSO. Thus, the EBPGA algorithm selects the best sentences covering all the topics of the original document to create the summary that is coherent, relevant, and nonredundant. The proposed algorithm was evaluated by producing summaries of size 20%, 30%, and 40%, of the original document, and compared with the existing algorithms. The proposed summarization algorithm obtained a Recall-Oriented Understudy for Gisting Evaluation (ROUGE) score that is comparable with the contemporary algorithms.

The traditional view-aware based vehicle re-identification methods integrated the vehicle view information into the training process, allowing the model parameters to learn the vehicle view knowledge. So, the extracted vehicle features contained vehicle view information, and the impact of view differences on model performance can be reduced. Since ensemble training with different methods will affect each other, it is difficult for these methods to train models together with other non view-aware methods. In order to address this challenge, in this paper, we propose a view-aware-based post-processing method (VABPP), which uses vehicle view information to re-rank the re-identification results during testing. According to vehicle views, VABPP method divide the distances between vehicle id features into several groups. During testing, multiply the distances of each different group by different coefficients. So, it treats different views equally. In order to achieve better implementation of this post-processing method, in this paper, we also propose a scheme that unifies the feature distance distributions of the training set and the test set. This scheme can enable some properties of the training set to be directly used in the test set. As the properties of the training set are trained under the guidance of correctly labeled labels, which enhances the robustness of the test set properties. The mAP in VeRi-776 dataset and in the three test sets of VERI-Wild dataset are 83.7%, 88.7%, 84.6% and 78.5%, respectively. And the rank-1 in the three test sets of VehicleID dataset are 88.6%, 85.4% and 81.1%, respectively.

Abstract Agriculture is an essential part of human life. The crop productivity is based on the soil and environmental factors. Different crops are cultivated in different areas. Nowadays, the crop productivity level is affected by the climate change and diseases in the crops. Due to this pest infestation, the crop growth is heavily affected. To overcome this problem, the right fertilizer for a particular crop has to be chosen and fertilizer helps farmers to improve the crop productivity rate. This process can be done by using various machine learning techniques. In this work, various features selection techniques with classifiers used to predict the suitable fertilizer for a crop. The experimental results show that recursive feature elimination along the proposed Heterogeneous Stacked Ensemble classifier gives better prediction rate than other methods.

It is essential to understand the underlying emotions that are imparted through speech in order to study social communications as well as to generate seamless human–computer interactions. Speech emotion recognition (SER) is a considerably challenging task due to the lack of sufficient data and the complex interdependence of phrases with the context and emotion they imply. This article presents RobinNet: a RoBERTa-and Inception-ResNet-V2-based novel multimodal network for SER. The model employs transfer learning to build two unimodal systems for text and audio features and then incorporates them into a single classifier through Intermediate Fusion. This work has been created after carefully analyzing the performance of various top-performing unimodal systems and then utilizing a fine-tuned RoBERTa-based model to represent the textual features. Furthermore, we utilize an Inception-ResNetV2 pretrained network for Speaker Identification and employ transfer learning to train it for the task of emotion recognition through speech using spectrogram augmentation. The proposed multimodal system combines the two modalities through intermediate fusion and gives out a weighted accuracy (WA) of 72.8% when evaluated against the interactive emotional dyadic motion capture (IEMOCAP) dataset. Experimental results reveal that the proposed multimodal system outperforms state-of-the-art (SOTA) solutions on the benchmark datasets IEMOCAP, multimodal emotion lines dataset (MELD), and CMU-MOSEI. The proposed model utilizes intermediate fusion unlike any of its predecessors that perform late fusion after significant independent processing, thereby improving the overall artificial multimodal representations.

Wireless sensor network (WSN) is a rapid surging technology promising many fruitful innovations for the user community. The use of WSNs for continuous monitoring of environmental factors in severe situations, such as detection of volcanoes, forest fires, floods, and so on. Despite WSN’s various potential applications, energy and deployment are two major concerns influencing the network life span. The primary requirement is to monitor node energy consumption in order to improve network performance. Many approaches are proposed in recent times for energy efficiency, this article introduces an associated hypergraph with Lie algebra of upper triangular matrices for energy efficient clustering and routing. Initially, the sensor nodes are formulated as a hypergraph, and each hyperedge is treated as a cluster, from which cluster-head (CH) selection is exploited by minimum hypergraph transversal. The routing decision is employed by constructing the upper triangular matrix (UTM), and Lie commutators of the UTM Lie algebra identifies the best relay nodes for data forwarding. The performance of the proposed work is evaluated using simulation results which shows the effectiveness of the scheme over compared protocols in terms of, number of alive nodes, energy consumption and number of packets received in BS.

Summary Security in wireless sensor networks (WSNs) is challenging, owing to resource, computing, and environmental constraints. Researchers have proposed several security systems for multicasting in WSNs, but none have offered foolproof security. To make high‐security multicasting a reality, the proposed I‐RLNMCDS‐ODMRP protocol uses state‐of‐the‐art techniques such as identity‐based cryptography, random linear network coding, the minimum connected dominating set, and the on‐demand multicast routing protocol. Also, this article analyzes computation and memory overhead in the proposed protocol in terms of time in seconds and bytes, respectively, and examines its security properties. In the experimental results, the proposed protocol does not take much time to compute encryption and decryption parameters. Further, it requires little memory to store multicast message information. In the future, the proposed protocol will be tested on healthcare applications for intrabody communication through wireless body area networks.

Unit electricity pricing is of vital importance in an electric grid network. It is essential to charge the customers in a fair manner. Traditional pricing models are found to be inadequate in the ability to charge customers fairly due to a lack of support for real-time communication between customers and electricity providers. With the introduction of smart devices in the electric grid domain, the real-time gathering of information is a seamless process. Such an electric network that uses smart devices is called a smart grid. In a smart grid network, electricity providers can monitor the electricity usage pattern of customers in a real-time manner, which can then be analyzed to determine the appropriate prices. To analyze the customer’s history of usage and price the electricity in a real-time manner, the computation must be performed with minimal latencies. Adoption of a fog computing layer in the smart grids can aid in the attainment of this goal. In this article, we propose a novel method for the pricing of electricity. In our approach, the electric demand of a household is predicted based on their past usage patterns. Users are then clustered into different bins based on their demands, and an evolutionary algorithm is used to generate the prices for the users present in different bins in a real-time manner to ensure the maximum attainable profit to a service provider.

Center loss is widely used as a supervision tool in deep learning method. However, the center loss also has some shortcomings, the most important of which is that it must be combined with softmax loss to run well. In this article, we sum up five shortcomings of center loss and solve all of them by proposing a dual distance center loss (DDCL). Compared with center loss, DDCL can run without the combination of softmax to supervise training the model. In addition, we verify the inconsistency between the proposed DDCL and softmax loss in the feature space. To be specifically, we add the Pearson distance on the basis of the Euclidean distance to the same center, which makes all features of the same class be confined to the intersection of a hypersphere and a hypercone in the feature space, strengthens the intraclass compactness of the center loss, and enhances the generalization ability of center loss. Moreover, by designing a Euclidean distance threshold between all center pairs, we not only strengthen the interclass separability of center loss, but also make the center loss (or DDCL) works well without the combination of softmax loss. We verify the effectiveness of DDCL in four datasets, two of which are widely used in the field of vehicle re-identification named VeRi-776 dataset and VehicleID dataset, and two other datasets are widely used in the field of person re-identification named Market1501 dataset and MSMT17 dataset. The experimental results of the proposed DDCL exceed that of the softmax loss in all the four datasets, indicating that our proposed method not only can run without the combination of softmax, but also has a high accuracy.

Green computing refers to sustainable, environment-friendly computing that harnesses information and technology. Green computing can be thought of as applying the principles of manufacturing and design to the use and disposal of electronic products (computers, printers, servers, mobile phones, and storage devices), so the environment is not impacted. The goals of green computing include the minimal use of electronic products that do not affect the surroundings, reduced use of hazardous substances, increased energy efficiency during the product’s lifetime, and advanced recycling/biodegradability of nonfunctional products and factory wastes. The objective of this article is to present the importance of green business and the operational analysis of a data center.

Crop cultivation prediction is an integral part of agriculture and is primarily based on factors such as soil, environmental features like rainfall and temperature, and the quantum of fertilizer used, particularly nitrogen and phosphorus. These factors, however, vary from region to region: consequently, farmers are unable to cultivate similar crops in every region. This is where machine learning (ML) techniques step in to help find the most suitable crops for a particular region, thus assisting farmers a great deal in crop prediction. The feature selection (FS) facet of ML is a major component in the selection of key features for a particular region and keeps the crop prediction process constantly upgraded. This work proposes a novel FS approach called modified recursive feature elimination (MRFE) to select appropriate features from a data set for crop prediction. The proposed MRFE technique selects and ranks salient features using a ranking method. The experimental results show that the MRFE method selects the most accurate features, while the bagging technique helps accurately predict a suitable crop. The performance of proposed MRFE technique is evaluated by various metrics such as accuracy (ACC), precision, recall, specificity, F1 score, area under the curve, mean absolute error, and log loss. From the performance analysis, it is justified that the MRFE technique performs well with 95% ACC than other FS methods.

The focus of the current research is on making computers as energy-efficient as possible, and applying innovative ideas to energy-related computer technology. It is anticipated that green information technology (IT) will rapidly become a reality and official organizational policy. Thus, green IT is not merely restricted to environmental strategies but is concerned with the overall development of people and society as a whole. In this regard, collaboration is to be explored to optimize business. This article examines the importance of green computing in sectors such as IT, networking, industries, and corporations.

Machine learning (ML) is a futuristic concept, utilized as a tool for modeling real-world applications. Today, healthcare worldwide has drawn the attention of ML, with its ability to analyze huge data sets and convert information into clinical insights that aid physicians in disease diagnosis and treatment planning, leading to low cost, better outcomes, and greater patient satisfaction. One of such medical applications calling for ML interventions is human infertility; an issue addressed by a set of medical procedures termed assisted reproductive technology (ART). The ART success rate, however, is very low because it is affected by a number of variables. ML techniques are now applied to predict ART outcomes to find strategies for an improved success rate. The literature reviewed on the subject shows that most of the available ART models are static. If ML is to have a role in healthcare, it must take an incremental, rather than static, approach. Therefore, this research proposes a dynamic model for ART outcome prediction. The model is built by an ensemble of two incremental classifiers, namely, the instance-based (IB1) learner and averaged one-dependence estimators (A1DEs) updatable learner through voting. The performance of the proposed model is checked with other ensemble models and ART data sets to find that the former shows promise in ART outcome prediction.