Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Rain is a typical meteorological phenomenon that can significantly impair the functionality of outdoor computer vision systems, including autonomous navigation and surveillance. Depending on how far away the streaks are from the camera, they may appear differently in the images. One input image serves as the foundation for the majority of current rain removal techniques. However, estimating a trustworthy depth map for rain removal is challenging on a single image. To overcome these challenges, this research introduces a novel approach for rain streak removal utilizing generative adversarial networks (GANs). Leveraging the discriminative power of GANs, the proposed technique effectively distinguishes between rain streaks and clean image content, resulting in the generation of realistic, rain-free images. The workflow involves initial image pre-processing using a cross-guided bilateral filter for detail layer extraction. The rain streak removal is then executed through an improved de-rain GAN (DR_GAN), where the generator module is replaced with a dense bidirectional network with self-attention (Attn_DBNet). This integration incorporates DenseNet-121, bidirectional gated recurrent unit (BiGRU), and self-attention mechanisms, enhancing the overall performance of the rain streak removal process. The research further introduces chaotic logistic gazelle optimization (CL-G) for optimizing the loss function, addressing local optimal trapping issues through the incorporation of chaotic logistic mapping. With notable gains in the metrics, comparative analysis shows that the proposed method is superior to the state-of-the-art approaches. These successes demonstrate the usefulness and superiority of the proposed GAN-based rain streak removal method over dual CNN, QSAM-Net, and MGPDNet approaches, with significant percentage advantages over these networks. PubDate: 2025-03-17
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: The increasing reliance on digital signatures for secure authentication and verification necessitates advanced watermarking techniques to protect signature integrity. Transform-domain methods, including the discrete cosine transform (DCT) and the discrete wavelet transform (DWT), are proposed for their potential to balance robustness, imperceptibility, and recognition accuracy in online signature biometrics. This study explores multi-bit watermarking approaches applied to online handwritten signatures using the MCYT signature database. We investigate the effects of embedding multiple bits per sample with adjustable watermark strength ($$\alpha $$) in the DCT and DWT domains. The trade-offs between signal distortion, watermark extraction accuracy, and biometric recognition rates are systematically evaluated. Experimental results reveal that while increasing $$\alpha $$ enhances watermark robustness, it also leads to perceptible distortions in signature samples. We identify the minimum $$\alpha $$ thresholds required for error-free watermark extraction and analyze their impact on identification and verification performance. The proposed multi-bit embedding strategy in the DCT domain demonstrates a viable compromise between robustness and imperceptibility, maintaining acceptable biometric recognition rates. Transform-domain watermarking techniques provide a promising solution for secure and robust online signature biometrics. This study highlights the feasibility of incorporating multi-bit watermarking schemes with adjustable strength into online signature systems, enhancing security while preserving recognition accuracy. PubDate: 2025-03-15
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Distributed learning is an effective method for solving large-scale cognitively inspired online machine learning problems. However, frequent communications between nodes lead to expensive communication burden. Meanwhile, projection operations because of the constraints in decision variables cause a lot of computational cost. In order to address the above problems, this paper presents a communication-efficient distributed Frank-Wolfe online optimization method, which integrates the event-triggered mechanism into the distributed projection-free online optimization algorithm. Furthermore, we provide a rigorous theoretical analysis for the regret of the proposed algorithm. Finally, we verify the performance of the proposed method through a variety of numerical experiments. The theoretical results show that the regret reaches a sublinear growth of iterations for convex objective functions. The proposed algorithm outperforms the baseline methods on two datasets. Our research indicates that, in addition to reducing computational overhead, the event-triggered scheme has the potential to enhance the communication efficiency of distributed network system. PubDate: 2025-03-14
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Large language models (LLMs) have made significant advancements in natural language processing (NLP), impacting both academia and industry. Evaluating LLMs is crucial, as these models are developed for multiple NLP tasks. However, no single LLM has successfully fulfilled all tasks simultaneously, creating a research gap. This, in turn, leads to the identification of the most and least effective LLMs in real-world problems, presenting a multi-criteria decision-making (MCDM) challenge due to the diversity of evaluation tasks, task prioritization, data variability, and issues related to ranking and grading with binary data. While the three-way decision (3WD) approach based on MCDM methods can address this, it often leaves uncertainty as an open issue, highlighting a theoretical gap. To address this, the contribution of this study is the development of a new 3WD approach based on utility and dynamic localization transformational procedures within a circular q-rung orthopair fuzzy set (C-q-ROFS) for ranking and grading LLMs. The methodology includes (1) reformulating the fuzzy weighted zero inconsistency-based interrelationship process (FWZICbIP) using C-q-ROFS (C-q-ROFS–FWZICbIP method) to prioritize tasks and address weighting uncertainty; (2) formulating a decision matrix by intersecting LLMs with NLP tasks while applying utility and dynamic localization procedures to handle binary input issues; and (3) reformulating the conditional probabilities by opinion scores (CPOS) method within the C-q-ROFS context (C-q-ROFS–CPOS method) to determine decision thresholds for each LLM. This involves incorporating Bayesian decision theory under C-q-ROFS to establish decision thresholds for all LLMs, thereby enhancing the certainty and effectiveness of the grading process. Based on this, the 3WD approach is developed to offer a robust mechanism for ranking and grading LLMs. Forty LLMs were ranked and graded across 11 NLP tasks, with the findings showing that LLM14 demonstrated high efficacy, ranking in the positive region for nine σ values, but falling into the boundary region at σ = 0.05. Sensitivity and comparison analyses were conducted to evaluate the robustness and stability of the methodology. PubDate: 2025-03-07
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Gestalt theory laid the foundation for modern cognitive learning theory and emphasizes that the whole is greater than the sum of its parts, where similarity and proximity are two important principles. However, exploiting Gestalt theory to detect multiple salient objects remains challenging. In this paper, we propose a very simple yet efficient saliency model based on Gestalt theory, namely, the color similarity and spatial proximity (CSSP) model. It utilizes content-based image retrieval (CBIR) techniques to detect salient objects. The methodology has three important highlights: (1) a novel weighted distance is proposed to calculate spatial proximity. It can control spatial proximity within a certain range and detect salient objects robustly. (2) Two novel and efficient saliency scoring calculation methods are proposed under the framework of CBIR techniques, where color similarity and spatial proximity are used for image matching and the ordering of retrieved images. This enables the robust identification of multiple salient objects. (3) A very simple yet efficient integration method is proposed to combine saliency maps. Using this integration method, impurities around salient objects are greatly reduced, and their interiors are highlighted robustly. Experiments with several well-known benchmark datasets validate the performance of the CSSP model. The CSSP method resulted in fewer grey patches inside salient objects, and it is superior to many existing state-of-the-art methods. The detected salient regions were brighter, improving the effectiveness of multiple salient objects detection. In addition, the CSSP method can detect salient objects robustly even when they touch the image boundaries. It has demonstrated that modeling visual attention based on Gestalt theory is a novel, viable approach. PubDate: 2025-03-05
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: In recent years, Chinese word embeddings have attracted significant attention in the field of natural language processing (NLP). The complex structures and diverse influences of Chinese characters present distinct challenges for semantic representation. As a result, Chinese word embeddings are primarily investigated in conjunction with characters and their subcomponents. Previous research has demonstrated that word vectors frequently fail to capture the subtle semantics embedded within the complex structure of Chinese characters. Furthermore, they often neglect the varying contributions of subword information to semantics at different levels. To tackle these challenges, we present a weight-based word vector model that takes into account the internal structure of Chinese words at various levels. The model further categorizes the internal structure of Chinese words into six layers of subword information: words, characters, components, pinyin, strokes, and structures. The semantics of Chinese words can be derived by integrating the subword information from various layers. Moreover, the model considers the varying contributions of each subword layer to the semantics of Chinese words. It utilizes an attention mechanism to determine the weights between and within the subword layers, facilitating the comprehensive extraction of word semantics. The word-level subwords act as the attention mechanism query for subwords in other layers to learn semantic bias. Experimental results show that the proposed word vector model achieves enhancements in various evaluation metrics, such as word similarity, word analogy, text categorization, and case studies. PubDate: 2025-03-03
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Semi-supervised learning with a graph-based approach has gained prominence in machine learning, particularly in scenarios where labeling data involves substantial costs. Graph convolution networks (GCNs) have found widespread application in semi-supervised learning, predominantly on graph-structured data such as citation and social networks. However, a noticeable gap exists in the application of these methods to non-graph multi-view data, such as collections of images. In an effort to address this gap, we introduce two innovative deep semi-supervised multi-view classification models specifically tailored for non-graph data. Both models share a common architecture, leveraging GCNs and integrating a label smoothing constraint. The primary distinction lies in the construction of the consensus similarity graph. The first model directly reconstructs the consensus graph from different views using a specialized objective function designed for flexible graph-based semi-supervised classification. In contrast, the second model independently reconstructs individual graphs and subsequently adaptively merges them into a unified consensus graph. Our experiments encompass various multiple-view image datasets. The results consistently demonstrate the superior performance of our proposed approach compared to traditional fusion methods with GCNs. In this research, we present two approaches for tackling semi-supervised classification challenges involving multiple views. One method is named Semi-supervised Classification with a Unified Graph (SCUG), and the other is referred to as Semi-supervised Classification with a Fused Graph (SC-Fused). Both methods share a common semi-supervised classification process, utilizing the GCN framework and incorporating label smoothing. However, the key distinction lies in the construction of the similarity graph. Unlike traditional ad hoc graph construction approaches, our proposed methods, SCUG and SC-Fused, estimate the unified graph or individual graphs, respectively, alongside the labels. This results in more optimized graphs that benefit from data smoothing and the semi-supervised context. PubDate: 2025-03-01
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: With the rise of online adaptive learning, Knowledge Tracing (KT) has become an indispensable component of online education systems. KT assesses the knowledge level of each learner by tracing their learning activities. Managing cognitive load is crucial in the learners’ cognitive process; too low a load may lead to a lack of concentration, while excessively high cognitive load can impede information processing. In pursuit of an ideal learning model, this paper proposes the Cognitive Load-based Knowledge Tracing (CLKT) model. This model employs a Heterogeneous Cognitive Graph Convolutional Network (HCGCN) to extract learners’ knowledge representations and establish connections between learning tasks or instructional resources and learners, providing the model with interpretable learning path recommendations. By introducing the Attention Concentration (AC) mechanism, the model dynamically processes information and efficiently integrates it into learners’ knowledge structures to maintain an appropriate cognitive load level, thus maximizing effective learning. Experiments conducted on the ASSISTMENTS dataset, which contains real-world student interaction data from an online tutoring system, focus on studying the impact of different cognitive loads on the learning process. The experimental results delve into the effects of cognitive load on learner performance, ensuring that learners can engage in learning with appropriate pace and difficulty, thereby enhancing their learning outcomes. PubDate: 2025-03-01
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Multi-expert multi-criteria decision-making (MEMCDM) based on interval-set information is novel and valuable, and it already adopts an effective strategy of alternative queuing method (AQM), called AQM-IS. AQM-IS mainly relies on dissimilarity measures and possibility degrees of interval sets, and the two types of uncertainty measures have absolute-quantitative limitations on rough information extraction to imply improvement space. In this work, improved dissimilarity measures and possibility degrees of interval sets are constructed from a better perspective of relative quantization related to systematic structuring and statistical fusion, so improved AQM (called IAQM-IS) is established to advance MEMCDM by using the interval-set information transformation. As bases, relative dissimilarity measures are proposed to modify absolute dissimilarity measures for both interval-set pairs and families on closeness and deviation; thus, relevant internal relationships, mutual sizes, axiomatic properties, and illustrative examples are acquired. Aiming at interval-set information, improved AQM (i.e., IAQM-IS) is investigated for MEMCDM. Concretely, absolute dissimilarity measures are chosen to determine criterion weights based on judgement matrix and maximum deviation, and improved possibility degrees of interval sets are proposed by systematic likelihood characterizations and arithmetic mean combination; using the weight arithmetic mean of improved dissimilarity measures and possibility degrees, a more powerful index for sorting alternatives is generated to formulate IAQM-IS. For algorithmic evaluation, two assessment indices of decision rankings (called separability and goodness) are designed; accordingly, the two algorithms of MEMCDM — AQM-IS and IAQM-IS — are demonstrated and compared via both an applied examples of e-commerce platforms and six simulated experiments of public datasets, and thus the effectiveness and superiority of IAQM-IS are verified. In summary by the double-quantization technique, the improved dissimilarity measures and possibility degrees deepen uncertainty measures of interval-set information tables, and corresponding IAQM-IS has better decision performance than current AQM-IS in specific application scenarios of social cognition. PubDate: 2025-02-28
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Emotional support conversation aims to convey understanding, sympathy, care, and support through conversation, to help others cope with emotional distress, pressure, or challenges. In this study, we conduct a holistic comparative study to investigate how well the most recent large language models (LLMs), which have recently proved to have empathy, and act as emotional supporters. To this end, we make use of the emotional support conversation (ESC) framework and assess multiple maintain LLMs accordingly. We then have an in-depth comparison between these LLM-based emotional supporters to humans in terms of the use of emotional support strategies as well as the use of language. Surprisingly, we find that there is still a huge gap until these LLMs become effective emotional supporters. This is because, on the one hand, they have strong preference biases on using a limited set of strategies, making them always show empathy but rarely take real actions (such as providing suggestions), which is key in ESC. On the other hand, they often over-generate responses, making what they utter a departure from those of human experts. PubDate: 2025-02-27
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Pneumoconiosis is a serious occupational disease with high morbidity and disability rates. However, the evolution of pneumoconiosis is complex and changeable, and the course of most cases is incomplete, resulting in a lack of continuity in a large number of data samples, which makes it challenging for radiologists to accurately assess the development of the disease. We propose a conditional self-attention TimesNet as the backbone network for time series analysis tasks, aiming to improve prognosis prediction based on disease progression information in pneumoconiosis image data at different time periods. In our approach, we train the model using chest X-ray images of the same patient at different time points, incorporating a hierarchical attention structure and self-attention blocks to fully consider the contextual correlation information of consecutive time-point images. Additionally, diverse clinical features of patients are utilized as conditional inputs in disease progression prediction. The goal is to better learn the progression status of the disease and reflect the disease trajectory representation of missing time series data, enhancing the model’s predictive capabilities. Simultaneously, an adversarial diffusion generation model is designed to fill in missing values in the time series data. The missing data generated by the model effectively improves radiologists’ judgment of pneumoconiosis progression. We trained our model using missing time series images to predict clinical outcomes. Experimental validation on two medical datasets shows that the AUC, sensitivity, specificity, and DSC achieved 90.33%, 87.89%, 85.01%, and 88.54%, respectively. These results highlight the competitive performance of our method across multiple evaluation metrics. Our model can capture the correlation between short-term/long-term/missing time series lesion features and time in pneumoconiosis images. This approach holds significant implications for predicting clinical outcomes and progression risk in pneumoconiosis, providing valuable guidance for the assessment and prognosis of pneumoconiosis. PubDate: 2025-02-25
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Traditional transformers struggle to effectively capture local contextual information. Conversely, CNNs face challenges in modeling long-range dependencies. To address these limitations, this paper introduces GSAC-UFormer, an innovative Groupwise Self-Attention Convolutional Transformer-based UNet for medical image segmentation. The design of GSAC-UFormer focuses on efficiently integrating both local and global information, balancing the strengths of different processing techniques. At the core of GSAC-UFormer is the GSAC-Former block. This module combines groupwise convolution with a CNN-adaptive self-attention mechanism, enabling parallel integration of local and global contexts. This architecture allows the model to effectively capture intricate dependencies across various data dimensions while processing local features with high efficiency. The Guided Contextual Feature Attention (GCFA) mechanism further enhances feature selection. It emphasizes the most relevant contextual information, refining spatial and channel-wise relationships in the extracted features. This targeted approach mitigates noise and improves model accuracy. Additionally, the Multi-Depth Partitioned Depthwise Convolution Transformer (MDPDC-Former) serves as a bottleneck module. It optimizes feature mapping and enhances network learning efficiency by dynamically adjusting the receptive field. This enables the model to capture multi-scale semantic information more effectively. Experimental results highlight the superior performance of GSAC-UFormer compared to state-of-the-art methods. It achieves Dice coefficients of 91.6%, 94.61%, and 82.24% on the MICCAI 2017 (red lesion), PH2, and CVC-ClinicalDB datasets, respectively. These results underscore its effectiveness in advancing medical image segmentation. PubDate: 2025-02-22
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Transfer learning–based fault diagnosis has received intensive attention from researchers. Under various working conditions, high-precision cross-domain fault diagnosis remains a problem due to distribution differences between different source domains and between source and target domains. Therefore, reducing the distribution difference between source domain and target domain data is crucial for improving the model’s ability to learn domain-invariant features and fault-representative features. To address this challenge, this paper proposes a multi-source adversarial domain adaptation approach for fault diagnosis, referred to as MSD-MCA, which is based on the alignment of multiple classifiers. The method constructs a sub-network for each source domain and utilizes domain adversarial training to extract domain-invariant features. It then generates a fault feature set for each source domain by leveraging the domain-invariant features corresponding to various fault types. To align the target domain with the source domains, the Wasserstein distance is calculated between the target features and each fault feature set. Minimizing the entropy of the distribution distance vector facilitates the learning of fault-representative features. Additionally, an association matrix is employed to enhance the stability of the decision boundaries during the training process. This approach improves the model’s capacity to generalize across multiple domains while effectively capturing fault-related information. To validate the efficacy of the proposed MSD-MCA method, a comparative analysis was conducted against several state-of-the-art diagnostic approaches. The evaluation was performed on bearing fault data from Case Western Reserve University, as well as two real-world industrial datasets. The results indicate that MSD-MCA shows improved accuracy and enhanced generalization capabilities across both datasets. Consequently, MSD-MCA can better learn the domain-invariant features and fault-representative features and improve the accuracy of fault diagnosis. PubDate: 2025-02-19
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Cancer is the second leading cause of death worldwide, and within this type of disease, lung cancer is the second most diagnosed, but the leading cause of death. Early detection is crucial to increase patient survival rates. One of the primary methods for detecting this disease is through medical imaging, which, due to its features, is well-suited for analysis by deep learning techniques. These techniques have demonstrated exceptional results in similar tasks. Therefore, this paper focusses on analyzing the latest work related to lung cancer detection using deep learning, providing a clear overview of the state of the art and the most common research directions pursued by researchers. We have reviewed DL techniques for lung cancer detection between 2018 and 2023, analyzing the different datasets that have been used in this domain and providing an analysis between the different investigations. In this state-of-the-art review, we describe the main datasets used in this field and the primary deep learning techniques used to detect radiological signs, predominantly convolutional neural networks (CNNs). As the impact of these systems in medicine can pose risks to patients, we also examine the extent to which explainable AI techniques have been applied to enhance the understanding of these systems, a crucial aspect for their real-world application. Finally, we will discuss the trends that the domain is expected to follow in the coming years and the challenges that researchers will need to address. PubDate: 2025-02-17
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Multi-attribute group decision-making (MAGDM) refers to a series of decision-making problems that rank all possible alternatives based on decision makers’ cognition and evaluations over alternatives from multiple attributes. Hence, the precondition of MAGDM is felicitously describing decision makers’ fuzzy and uncertain cognitive information in complicated decision-making issues. The recently proposed linguistic q-rung orthopair fuzzy set (Lq-ROFS), which uses two linguistic terms to denote membership and non-membership degrees, has been proved to be an effective and promising tool to depict decision makers’ complex cognition in real MAGDM problems. Considering the drawbacks of existing Lq-ROFS-based decision-making methods, this paper focuses on MAGDM approaches where decision makers’ cognitive information is denoted by Lq-ROFSs. The main contribution of this paper is to propose a novel MAGDM method based on Lq-ROFSs. This paper introduces a new MAGDM method under Lq-ROFSs. In order to do this, this study first puts forward some new operational rules for linguistic q-rung orthopair fuzzy numbers (Lq-ROFNs) based on Archimedean copula. These new operational rules are more flexible than existing ones and some other operations can be derived by using different generators. Second, to effectively aggregate Lq-ROFNs, the extended power average operator is applied in linguistic q-rung orthopair fuzzy environment and based on the new operational rules, some novel aggregation operators are generated. Afterward, the developed aggregation operators are used in decision-making problems and a novel MAGDM method which concentrates on linguistic q-rung orthopair fuzzy decision environment is introduced. Specific steps of the new method are illustrated in detail and it is then applied in some illustrative examples to verify its effectiveness. Our proposed method is effective for handling MAGDM problems under Lq-ROFSs. Numerical examples have shown the effectiveness in handling realistic MAGDM problems. In addition, comparison with some existing methods illustrates the advantages and superiorities of our method. This paper introduces a new MAGDM method under Lq-ROFSs. This method is based on Archimedean copula, extended power average operator, and Lq-ROFSs, and is powerful and flexible to cope with MAGDM problems in reality. PubDate: 2025-02-14
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: With the prosperity of social media, toxic language spreading over social media has become an unignorable challenge for individual mental health and social harmony. Many researchers have studied toxic language identification to control or mitigate it. However, it still leaves a blank in the cognitive patterns of toxic language. Metaphors as a common feature in natural language connect literal and metaphorical meanings, which could be a useful tool to study the underlying cognitive patterns of the text. In this paper, we utilize a metaphor processing tool, MetaPro, to process a public toxic language dataset and analyze the cognitive biases between toxic and non-toxic language, multiple levels and subtypes of toxic language as well as toxic language mentioning different genders, sexual orientations, and races. Our study demonstrates that significant differences exist in cognitive patterns of the above-mentioned categories and analyzes the differences with machine learning methods. PubDate: 2025-02-14
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Cognitive processes that affect people perceptions, comprehension, and interactions with others, such as attribution mistakes and heuristics, are responsible for social issues. These social issues includes polarization, prejudice, and inequality. To address these issues, we must comprehend cognitive mechanisms, and this can be made by using some appropriate multi-attribute decision-making (MADM) approach, that can handle people perceptions of complex and bipolar nature. Thus, in this manuscript, we concentrate on a MADM technique that relies on certain novel aggregation operators in the framework of bipolar complex fuzzy sets. These aggregation operators include Muirhead mean (MM) operator and dual Muirhead mean (DMM) operator of several types. To authenticate the validity of these defined aggregation operators, certain properties of these operators are proved. Furthermore, we consider the interpreted operators to produce a decision-making (DM) technique to deal with bipolar complex fuzzy MADM issues. We then consider a real life example to show the application and need of the interpreted work in daily life. To confirm the viability and potential of the offered technique, we compare our established technique with some other prevailing techniques. PubDate: 2025-02-07
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Generative artificial intelligence (GAI) represents a pioneering class of artificial intelligence systems renowned for producing diverse media, such as text and images. Agriculture 4.0 (AG-4.0) is a concept that integrates advanced technologies such as the Internet of Things (IoT), data analytics, artificial intelligence, and precision agriculture into the agricultural sector. The integration of GAI and AG-4.0 can generate new and valuable agricultural insights and solutions through pattern recognition and data analysis. This integration enhances farming practices by generating predictive models, simulating optimal growth conditions, diagnosing plant diseases, and optimizing genetic traits. In spite of the tremendous scope of GAI in agriculture, there has been no detailed study concerning the applications and scope of GAI in AG-4.0. Addressing this research gap, we explore various applications, real-world products, and limitations of GAI in agriculture. We explore how GAI models such as ChatGPT and Dall-E can be personalized advisors for farmers, help increase awareness about farmer relief programs, design farm layouts, and many other such applications. Additionally, we cover four real-world GAI products deployed to assist farmers. Since GAI is a growing technology, it poses challenges such as scarcity of data, data privacy, and interpretability. We elaborately discuss these limitations and suggest multiple directions for future research in GAI for agriculture. PubDate: 2025-02-04
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Decision-making is a complex cognitive process, in which emotion is one of the most important factors. But insights into the influence of emotion on decision-making are scarce, especially the underlying mechanism of the brain. To reveal the brain’s underlying mechanisms of the influence of emotion on decision-making, an experiment involving emotion elicitation and decision-making tasks was designed. Electroencephalography (EEG), behavioral, and subjective data were collected and conducted. We constructed time-varying weighted directed networks by phase slope index (PSI) in four frequency bands and calculated graph theory metrics. Firstly, the period that the brain processes information most efficiently is 100–300 ms after the appearance of the decision-making task. Secondly, by analyzing the temporal-spatial domains of EEG, the significant differences in global efficiency (GE) and local efficiency (LE) were found among three different emotion groups in the alpha band in the low-difficulty task during 100–300 ms. Thirdly, most activation regions of different emotions were similar and concentrated in the parietal, and occipital lobes but there were still slight differences that were more likely to be found in the prefrontal and left temporal lobes. Graph theory metrics in the decision-making process changed dynamically in the temporal domain and graph theory metrics of different emotions were different. PubDate: 2025-02-01
Please help us test our new pre-print finding feature by giving the pre-print link a rating. A 5 star rating indicates the linked pre-print has the exact same content as the published article.
Abstract: Forecasting cryptocurrency prices is challenging due to market volatility and dynamic behavior. This study aims to enhance prediction accuracy for Bitcoin (BTC), Ethereum (ETH), and Litecoin (LTC) by proposing a novel deep learning framework. The framework integrates the Sparrow Search Algorithm (SSA) for selecting optimal technical indicators with Bidirectional Long Short-Term Memory (Bi-LSTM) networks. Technical indicators derived from historical market data, including prices and trading volume, are analyzed to improve forecasting. The results demonstrate that the proposed framework effectively enhances prediction accuracy for BTC and LTC. For ETH, the best performance is achieved using all 34 indicators with the Bi-LSTM model. These findings highlight the importance of selecting relevant indicators and demonstrate the potential of advanced deep learning models in addressing the complexities of cryptocurrency markets. This research provides valuable insights and a reliable framework for improving cryptocurrency price predictions. PubDate: 2025-02-01
Hybrid journal (It can contain Open Access articles)
[2468 journals]
