Multiple Networks Research Articles

Enhancing object recognition: The role of object knowledge decomposition and component-labeled datasets

Deep learning models’ decision-making processes can be elusive, often raising concerns about their reliability. To address this, we have introduced the Object Knowledge Decomposition and Components Label Dataset (OKD-CL), designed to improve the interpretability and accuracy of object recognition models. This dataset includes 99 categories from PartImageNet, each detailed with clear physical structures that align with human visual concepts. In a hierarchical structure, every category is described by Abstract Component Knowledge (ACK) descriptions and each image instance comes with Explicit Visual Knowledge (EVK) masks, highlighting the visual components’ appearance. By evaluating multiple deep neural networks guided with ACK and EVK (dual-knowledge-guidance approach), we saw better accuracy and a higher Foreground Reasoning Ratio (FRR), confirming our knowledge-guided method’s effectiveness. When used on the Hard-ImageNet dataset, this approach reduced the model’s reliance on incorrect feature assumptions without sacrificing classification accuracy. This hierarchical comprehension encouraged by OKD-CL is crucial in minimizing incorrect feature associations and strengthening model robustness. The entire code and dataset are available on: https://github.com/XiGuaBo/OKD-CL.

Characterization of changes in the resting-state intrinsic network in patients with diabetic peripheral neuropathy

Diabetic peripheral neuropathy (DPN) is the most common complication of type 2 diabetes mellitus (T2DM) and is often accompanied by a variety of cognitive and emotional deficits, but the neurologic mechanisms underlying these deficits have not been fully elucidated. Therefore, this study aimed to use independent component analysis to explore the changes in the characteristics within the intrinsic network and to reveal patterns of interactions between networks in patients with DPN. Forty-one patients with T2DM who showed DPN, 37 patients with T2DM who did not show DPN (NDPN group), and 43 healthy controls (HC) underwent a neuropsychological assessment and resting-state functional magnetic resonance imaging examinations to examine the patterns of intra- and inter-network variations in the patients with T2DM at different clinical stages (with and without DPN). The relationships of intra- and inter-network functional connectivity (FC) with clinical/cognitive variables were also examined. In comparison with the NDPN group and HC, patients with DPN showed decreased FC within the visual network and sensorimotor network (SMN). Moreover, in comparison with the HC group, patients with DPN showed decreased FC within the anterior default mode network and increased FC within the basal ganglia network. Inter-network analysis showed decreased FC between the SMN and salience network in patients with DPN relative to the NDPN and HC groups. The decreased FC within the bilateral paracentral lobule (BA 6) of SMN was associated with Color Trails Test part 1 scores (r = -0.302, P = 0.007) and disease duration (r = -0.328, P = 0.003) in all patients with T2DM. In conclusion, the results revealed that patients with DPN have abnormal FC in multiple resting-state intrinsic networks in addition to the SMN, and that decreased FC between the SMN and salience network may be involved in the neural basis of abnormal sensorimotor function in patients with DPN.

RNDiff: Rainfall nowcasting with Condition Diffusion Model

The Diffusion Models are widely used in image generation because they can generate high-quality and realistic samples. In contrast, generative adversarial networks (GANs) and variational autoencoders (VAEs) have some limitations in terms of image quality. We introduce a diffusion model to the precipitation forecasting task and propose a short-term precipitation nowcasting with condition diffusion model based on historical observational data, which is referred to as Rainfall nowcasting with Condition Diffusion Model(RNDiff). By incorporating an additional conditional decoder module in the denoising process, RNDiff achieves end-to-end conditional rainfall prediction. RNDiff is composed of two networks: a denoising network and a conditional encoder network. The conditional network is composed of multiple independent UNet networks. These networks extract conditional feature maps at different resolutions, providing accurate conditional information that guides the diffusion model for conditional generation. RNDiff surpasses GANs in terms of prediction accuracy, although it requires more computational resources. The RNDiff model exhibits higher stability and efficiency during training than GANs-based approaches, and generates high-quality precipitation distribution samples that better reflect future actual precipitation conditions. Compared to the current state-of-the-art GAN-based methods, our proposed approach achieves significant improvements on key evaluation metrics. Specifically, our method leads to improvements in the CSI, HSS, and FSS, which are increased by around 8%, 5%, and 6%, respectively. The experiment fully verified the advantages and potential of RNdiff in precipitation forecasting and provided new insights for improving rainfall forecasting. Our project is open source and available on GitHub at: https://github.com/ybu-lxd/RNDiff.

Decoding threat neurocircuitry representations during traumatic memory recall in PTSD.

The neurocircuitry mechanisms underlying recall of traumatic memories remain unclear. This study investigated whether traumatic memory recall engages neurocircuitry representations that mirror activity patterns engaged during generalized threat stimulus processing in Post Traumatic Stress Disorder (PTSD). Multivariate pattern analysis was used to train 3 decoders. A "trauma" decoder was trained on fMRI patterns during idiographic trauma versus neutral narratives in a sample of 73 adult women with PTSD. A separate cohort of 125 adult participants completed a reward and threat learning task, from which "shock" and "reward loss" decoders were trained on neural patterns during threat or reward outcome delivery, respectively. These decoders were then cross-tested on the alternative datasets, allowing analyses of the degree to which traumatic memory recall engaged neurocircuitry representations that overlap with more general aversive stimuli. Decoders were trained and tested in four networks related to salience processing as well bilateral amygdala and hippocampal masks. The shock decoder trained in a midcingulate / posterior insula network demonstrated elevated predictions for shock during traumatic versus neutral memory recall. Similarly, the trauma decoder made elevated predictions about trauma recall during shock versus no shock delivery across multiple networks related to salience processing. There was no overlap between reward loss decoder predictions and trauma memory recall or vice versa. PTSD participants with elevated re-experiencing symptoms demonstrated the highest engagement of shock activity patterns during trauma memory recall. These results suggest that trauma memory recall engages neurocircuitry representations that overlap with threat, specifically painful, stimulus delivery.

Effects of experimental nutrient enrichment and eutrophication on microbial community structure and function in “marine lakes”

Biogeochemical cycling of nitrogen (N) and carbon (C) are regulated by microbial communities that can respond to altered N and C availability, yet the form, consistency, and magnitude of these responses remain poorly constrained. We enriched and eutrophied distinct microbial communities in “marine lakes” through experimental additions of N (5 μM NH4Cl), C (100 μM sucrose), and combined N+C (same forms and concentrations), examining microbial community responses via 16S rRNA sequencing in conjunction with functional responses represented by net community production (NCP) and community respiration (CR) measurements. Individual N or C additions drove significant shifts in microbial community structure only sporadically, and rarely with a corresponding difference in NCP or CR rates. In contrast, the combined addition of N+C elicited strong coincident responses in microbial community structure and function: NCP and CR rates shifted sharply toward heterotrophy and were correlated with multiple microbial networks (r2 = 0.309–0.599, P < 0.001) that included globally distributed marine bacteria. Across multiple experiments, the consistent response of one network, comprised primarily of gammaproteobacterial heterotrophs (particularly Vibrio and Alteromonas), led initially dissimilar communities to converge toward similar composition. However, the distinct response patterns of other more diverse networks were superimposed on top of this network, indicating that inorganic N and organic C enrichment have multilayered effects on microbial communities. Collectively our results demonstrate that elevated N and C alter microbial community structure and function, selecting for multiple microbial networks that compete for, and rapidly cycle, N and C.

Active Learning with Particle Swarm Optimization for Enhanced Skin Cancer Classification Utilizing Deep CNN Models.

Skin cancer is a critical global health issue, with millions of non-melanoma and melanoma cases diagnosed annually. Early detection is essential to improving patient outcomes, yet traditional deep learning models for skin cancer classification are often limited by the need for large, annotated datasets and extensive computational resources. The aim of this study is to address these limitations by proposing an efficient skin cancer classification framework that integrates active learning (AL) with particle swarm optimization (PSO). The AL framework selectively identifies the most informative unlabeled instances for expert annotation, minimizing labeling costs while optimizing classifier performance. PSO, a nature-inspired metaheuristic algorithm, enhances the selection process within AL, ensuring the most relevant data points are chosen. This method was applied to train multiple Convolutional Neural Network (CNN) models on the HAM10000 skin lesion dataset. Experimental results demonstrate that the proposed AL-PSO approach significantly improves classification accuracy, with the Least Confidence strategy achieving approximately 89.4% accuracy while using only 40% of the labeled training data. This represents a substantial improvement over traditional approaches in terms of both accuracy and efficiency. The findings indicate that the integration of AL and PSO can accelerate the adoption of AI in clinical settings for skin cancer detection. The code for this study is publicly available at ( https://github.com/Sayantani-31/AL-PSO ).

An active queue management method for multiple bottleneck networks based on distributed average tracking

The active queue management (AQM) is a key technology for the information infrastructure and the Internet. The AQM for multiple bottleneck networks is especially challenging than that for single-bottleneck networks because the stability analysis of multiple bottleneck networks is more complex. In this paper, the active queue management problem for multiple bottleneck networks is considered. A novel random early detection algorithm is designed based on distributed average tracking technology, which is used to estimate the global queue length and improve the network's performance. Sufficient conditions for global asymptotic stability of the closed-loop active queue management system are presented, and some simulation results with NS-2 are given. Compared with the RED algorithm, simulation results show that the presented algorithm has smaller oscillations at routers, which highlights its effectiveness.

A multiple physical crosslinked cellulose-based bioplastics with robust mechanical and thermal stability

The widespread use of traditional petroleum-based plastics has created an environmental crisis and health hazard, so there is an urgent need for bioplastics with excellent performance. However, fabricating of robust mechanical properties and heat resistance bioplastics in an efficient way has remained an enormous challenge. Herein, we proposed a strategy for the synergistic preparation of high-performance bioplastics with multiple physical crosslinking network structures via noncovalent and coordination bonds. In this strategy, carboxylated cellulose nanofibers (CNFs) and the polyphenol structures of tannic acid (TA) interacted noncovalently to create network structures; the bioplastic immersed in Ca2+ solution formed ionic crosslinked networks and TA-Ca coordination bonds. The synergistic effect of multiple network structures composed of hydrogen and coordination bonds made cellulose-based bioplastics have dense structures and robust tensile strength (114.2 MPa), while bioplastics had the characteristics of high transparency and superior thermal stability. Furthermore, the laminated composites formed by the bioplastic and PVA could support 1,000 g easily, which allowed it to be used as weighing application. Thus, the proposed multiple physical crosslinking strategy provides a method for developing cellulose-based bioplastics with excellent performance, which offers a new approach for the subsequent development of sustainable green materials.

Security enhancement for NOMA-PON with 2D cellular automata and Turing pattern cascading scramble aided fixed-point extended logistic chaotic encryption

The non-orthogonal multiple access-passive optical network (NOMA-PON) is facing the dual security threats of primary user interference and unauthorized third-party user eavesdropping, so efficient data security enhancement techniques are crucial. To solve these problems, we propose a fixed-point extended (FE)-logistic chaotic mapping to reduce the computational complexity while employing a two-dimensional (2D) cellular automata (CA) and Turing pattern (TP) cascading scramble (CA-TPCS) encryption algorithm to further improve the sensitivity of the NOMA-PON system. The CA-TPCS consists of 2D-CA dynamic bit encryption and Turing symbol substitution (TSS). By using FE chaos to construct 2D-CA and adopting index sort to extract the TSS matrix, dynamic diffusion of bits and scrambling of a 2D symbol matrix are achieved. To ensure the key privacy, we employ a dual key mechanism, and uplink data is introduced as the private key. To verify the feasibility of the proposed method, a simulation validation is built on a 17.6 Gb/s power division multiplexing-orthogonal frequency division multiplexing (PDM-OFDM) NOMA-PON system transmitted over 25 km standard single mode fiber (SSMF). The results show that the proposed scheme has no effect on the optimal power allocation rate (PAR) values and the values are all 3. Meanwhile, the receiver sensitivity gains of 0.2 and 0.3 dB are obtained for high-power and low-power users after encryption. The ciphertext has good diffusion and statistical properties, and the key space is flexibly controlled by the FE precision f, the length l of the transmitted bit, and the size T of the TP, with the value of 22f+l+T×T. The results show that the proposed scheme is not only very compatible with PDM technology but also can realize the dual defense of internal aggression and external aggression. It has a good application prospect in the future NOMA-PON.

Prediction of miRNA-disease association based on multisource inductive matrix completion

MicroRNAs (miRNAs) are endogenous non-coding RNAs approximately 23 nucleotides in length, playing significant roles in various cellular processes. Numerous studies have shown that miRNAs are involved in the regulation of many human diseases. Accurate prediction of miRNA-disease associations is crucial for early diagnosis, treatment, and prognosis assessment of diseases. In this paper, we propose the Autoencoder Inductive Matrix Completion (AEIMC) model to identify potential miRNA-disease associations. The model captures interaction features from multiple similarity networks, including miRNA functional similarity, miRNA sequence similarity, disease semantic similarity, disease ontology similarity, and Gaussian interaction kernel similarity between miRNAs and diseases. Autoencoders are used to extract more complex and abstract data representations, which are then input into the inductive matrix completion model for association prediction. The effectiveness of the model is validated through cross-validation, stratified threshold evaluation, and case studies, while ablation experiments further confirm the necessity of introducing sequence and ontology similarities for the first time.

Internal physiological drivers of leaf development in trees: Understanding the relationship between non‐structural carbohydrates and leaf phenology

Abstract Plant phenology is crucial for understanding plant growth and climate feedback. It affects canopy structure, surface albedo, and carbon and water fluxes. While the influence of environmental factors on phenology is well‐documented, the role of plant intrinsic factors, particularly internal physiological processes and their interaction with external conditions, has received less attention. Non‐structural carbohydrates (NSC), which include sugars and starch essential for growth, metabolism and osmotic regulation, serve as indicators of carbon availability in plants. NSC levels reflect the carbon balance between photosynthesis (source activity) and the demands of growth and respiration (sink activity), making them key physiological traits that potentially influence phenology during critical periods such as spring leaf‐out and autumn leaf senescence. However, the connections between NSC concentrations in various organs and phenological events are poorly understood. This review synthesizes current research on the relationship between leaf phenology and NSC dynamics. We qualitatively delineate seasonal NSC variations in deciduous and evergreen trees and propose testable hypotheses about how NSC may interact with phenological stages such as bud break and leaf senescence. We also discuss how seasonal variations in NSC levels, align with existing conceptual models of carbon allocation. Accurate characterization and simulation of NSC dynamics are crucial and should be incorporated into carbon allocation models. By comparing and reviewing the development of carbon allocation models, we highlight the shortcomings in current methodologies and recommend directions to address these gaps in future research. Understanding the relationship between NSC, source–sink relationships, and leaf phenology poses challenges due to the difficulty of characterizing NSC dynamics with high temporal resolution. We advocate for a multi‐scale approach that combines various methods, which include deepening our mechanistic understanding through manipulative experiments, integrating carbon sink and source data from multiple observational networks with carbon allocation models to better characterize the NSC dynamics, and quantifying the spatial pattern and temporal trends of the NSC‐phenology relationship using remote sensing and modelling. This will enhance our comprehension of how NSC dynamics impact leaf phenology across different scales and environments. Read the free Plain Language Summary for this article on the Journal blog.

Multi-Head Attention Affinity Diversity Sharing Network for Facial Expression Recognition

Facial expressions exhibit inherent similarities, variability, and complexity. In real-world scenarios, challenges such as partial occlusions, illumination changes, and individual differences further complicate the task of facial expression recognition (FER). To further improve the accuracy of FER, a Multi-head Attention Affinity and Diversity Sharing Network (MAADS) is proposed in this paper. MAADS comprises a Feature Discrimination Network (FDN), an Attention Distraction Network (ADN), and a Shared Fusion Network (SFN). To be specific, FDN first integrates attention weights into the objective function to capture the most discriminative features by using the proposed sparse affinity loss. Then, ADN employs multiple parallel attention networks to maximize diversity within spatial attention units and channel attention units, which guides the network to focus on distinct, non-overlapping facial regions. Finally, SFN deconstructs facial features into generic parts and unique parts, which allows the network to learn the distinctions between these features without having to relearn complete features from scratch. To validate the effectiveness of the proposed method, extensive experiments were conducted on several widely used in-the-wild datasets including RAF-DB, AffectNet-7, AffectNet-8, FERPlus, and SFEW. MAADS achieves the accuracy of 92.93%, 67.14%, 64.55%, 91.58%, and 62.41% on these datasets, respectively. The experimental results indicate that MAADS not only outperforms current state-of-the-art methods in recognition accuracy but also has a relatively low computational complexity.

Integrated energy microgrids participating in voltage regulation ancillary services: An improved ADMM based distributed optimization approach

AbstractThe integrated energy microgrid (IEM) has good potential for both active and reactive power regulation, which are gradually becoming critical resources for participating in power system ancillary services. This paper constructs a distributed optimization model for the IEMs participating in voltage regulation ancillary services by considering multiple physical network constraints. To reduce computational complexity, the DistFlow power flow model and the second‐order cone relaxation method are introduced to transform the optimization model into a second‐order cone programming problem. Moreover, an improved accelerated consensus alternating direction method of multipliers algorithm is proposed to accelerate distributed optimization solutions of the model while protecting user privacy. Finally, the proposed optimization model and solution method are tested in the modified IEEE‐33 node test case to verify their effectiveness and feasibility.

Distributed Passive Positioning and Sorting Method for Multi-Network Frequency-Hopping Time Division Multiple Access Signals.

When there are time division multiple access (TDMA) signals with large bandwidth, waveform aliasing, and fast frequency-hopping in space, current methods have difficulty achieving the accurate localization of radiation sources and signal-sorting from multiple network stations. To solve the above problems, a distributed passive positioning and network stations sorting method for broadband frequency-hopping signals based on two-level parameter estimation and joint clustering is proposed in this paper. Firstly, a two-stage filtering structure is designed to achieve control filtering for each frequency point. After narrowing down the parameter estimation range through adaptive threshold detection, the time difference of arrival (TDOA) and the velocity difference of arrival (VDOA) can be obtained via coherent accumulating based on the cross ambiguity function (CAF). Then, a multi-station positioning method based on the TDOA/VDOA is used to estimate the position of the target. Finally, the distributed joint eigenvectors of the multi-stations are constructed, and the signals belonging to different network stations are effectively classified using the improved K-means method. Numerical simulations indicate that the proposed method has a better positioning and sorting effect in low signal-to-noise (SNR) and low snapshot conditions compared with current methods.

Towards Integrating Federated Learning with Split Learning via Spatio-temporal Graph Framework for Brain Disease Prediction.

Functional Magnetic Resonance Imaging (fMRI) is used for extracting blood oxygen signals from brain regions to map brain functional connectivity for brain disease prediction. Despite its effectiveness, fMRI has not been widely used: on the one hand, collecting and labeling the data is time-consuming and costly, which limits the amount of valid data collected at a single healthcare site; on the other hand, integrating data from multiple sites is challenging due to data privacy restrictions. To address these issues, we propose a novel, integrated Federated learning and Split learning Spatio-temporal Graph framework (FS2G). Specifically, we introduce federated learning and split learning techniques to split a spatio-temporal model into a client temporal model and a server spatial model. In the client temporal model, we propose a time-aware mechanism to focus on changes in brain functional states and use an InceptionTime model to extract information about changes in the brain states of each subject. In the server spatial model, we propose a united graph convolutional network to integrate multiple graph convolutional networks. Integrating federated learning and split learning, FS2G can utilize multi-site fMRI data without violating data privacy protection and reduce the risk of overfitting as it is capable of learning from limited training data sets. Moreover, it boosts the extraction of spatio-temporal features of fMRI using spatio-temporal graph networks. Experiments on ABIDE and ADHD200 datasets demonstrate that our proposed method outperforms state-of-the-art methods. In addition, we explore biomarkers associated with brain disease prediction using community discovery algorithms using intermediate results of FS2G. The source code is available at https://github.com/yutian0315/FS2G.

Change in striatal functional connectivity networks across 2 years due to stimulant exposure in childhood ADHD: results from the ABCD sample.

Widely prescribed for Attention-Deficit/Hyperactivity Disorder (ADHD), stimulants (e.g., methylphenidate) have been studied for their chronic effects on the brain in prospective designs controlling dosage and adherence. While controlled approaches are essential, they do not approximate real-world stimulant exposure contexts where medication interruptions, dosage non-compliance, and polypharmacy are common. Brain changes in real-world conditions are largely unexplored. To fill this gap, we capitalized on the observational design of the Adolescent Brain Cognitive Development (ABCD) study to examine effects of stimulants on large-scale bilateral cortical networks' resting-state functional connectivity (rs-FC) with 6 striatal regions (left and right caudate, putamen, and nucleus accumbens) across two years in children with ADHD. Bayesian hierarchical regressions revealed associations between stimulant exposure and change in rs-FC of multiple striatal-cortical networks, affiliated with executive and visuo-motor control, which were not driven by general psychotropic medication. Of these connections, three were selective to stimulants versus stimulant naive: reduced rs-FC between caudate and frontoparietal network, and between putamen and frontoparietal and visual networks. Comparison with typically developing children in the ABCD sample revealed stronger rs-FC reduction in stimulant-exposed children for putamen and frontoparietal and visual networks, suggesting a normalizing effect of stimulants. 14% of stimulant-exposed children demonstrated reliable reduction in ADHD symptoms, and were distinguished by stronger rs-FC reduction between right putamen and visual network. Thus, stimulant exposure for a two-year period under real-world conditions modulated striatal-cortical functional networks broadly, had a normalizing effect on a subset of networks, and was associated with potential therapeutic effects involving visual attentional control.

How Freely Moving Mind Wandering Relates to Creativity: Behavioral and Neural Evidence.

Background: Previous studies have demonstrated that mind wandering during incubation phases enhances post-incubation creative performance. Recent empirical evidence, however, has highlighted a specific form of mind wandering closely related to creativity, termed freely moving mind wandering (FMMW). In this study, we examined the behavioral and neural associations between FMMW and creativity. Methods: We initially validated a questionnaire measuring FMMW by comparing its results with those from the Sustained Attention to Response Task (SART). Data were collected from 1316 participants who completed resting-state fMRI scans, the FMMW questionnaire, and creative tasks. Correlation analysis and Bayes factors indicated that FMMW was associated with creative thinking (AUT). To elucidate the neural mechanisms underlying the relationship between FMMW and creativity, Hidden Markov Models (HMM) were employed to analyze the temporal dynamics of the resting-state fMRI data. Results: Our findings indicated that brain dynamics associated with FMMW involve integration within multiple networks and between networks (r = -0.11, pFDR < 0.05). The links between brain dynamics associated with FMMW and creativity were mediated by FMMW (c' = 0.01, [-0.0181, -0.0029]). Conclusions: These findings demonstrate the relationship between FMMW and creativity, offering insights into the neural mechanisms underpinning this relationship.

Scalable production of robust and creep resistant ultra-high filled wood-plastic composites

With the widespread use of wood-based materials in human life, the availability of wood resources has gradually decreased. The use of low-value wood that does not require chemical adhesives can address the depletion of wood resources used to prepare wood-based composites. However, the development of high-strength, low-cost, scalable wood-based composites from low-value wood is challenging. In this study, high-performance ultra-high filled wood-plastic composites (UFWPC) composed of up to 95 wt% wood flour were prepared through cell wall densification and the construction of multiple cross-linked networks via deep cross-fusion. The UFWPC exhibited excellent mechanical properties, with a flexural strength that was 5.9 times higher than that of commercial particleboard, 2.1 times higher than commercial fiberboard, and 2.6 times higher than commercial wood-plastic composites. UFWPC also demonstrated excellent creep resistance, with a creep strain 76.79 % lower than that of commercial wood-plastic composites. Finally, a customizable large-scale commercial continuous flat-pressing system was established to produce UFWPC. The highly efficient preparation of UFWPC makes it an excellent alternative to commercial wood-plastic composites, particleboard, and fiberboard. This approach provides a promising valorization and sustainability method for recycling plastics and low-value wood.

Leveraging Mixture of Experts and Deep Learning-Based Data Rebalancing to Improve Credit Fraud Detection

Credit card fraud detection is a critical challenge in the financial sector due to the rapidly evolving tactics of fraudsters and the significant class imbalance betweenegitimate and fraudulent transactions. Traditional models, while effective to some extent, often suffer from high false positive rates and fail to generalize well to emerging fraud patterns. In this paper, we propose a novel approach that integrates a Mixture of Experts (MoE) model with a Deep Neural Network-based Synthetic Minority Over-sampling Technique (DNN-SMOTE) to enhance fraud detection performance. The MoE modeleverages multiple specialized expert networks, each trained to detect specific types of fraud, while the DNN-SMOTE generates high-quality synthetic samples to address the class imbalance. Our experimental results on a publicly available dataset demonstrate that the proposed method achieves a classification accuracy of 99.93%, a true positive rate of 84.69%, and a true negative rate of 99.95%. The Matthews Correlation Coefficient (MCC) of 0.7883 further highlights the model’s balanced performance in detecting fraudulent transactions. These results underscore the effectiveness of combining MoE with DNN-SMOTE, offering a robust solution for real-world credit card fraud detection scenarios.

Sediment Simulation Using Multi Layer Perceptron (MLP), Co-Active Neuro-Fuzzy Inference System (CANFIS) and Multiple Linear Regression Tecniques (MLR) for Hurdag Watershed

Sediment modeling plays a crucial role in sustainable water resources planning, development, and management. Techniques like the Multilayer Perceptron (MLP), Co-active Neuro-Fuzzy Inference System (CANFIS), and Multiple Linear Regression (MLR) have proven effective for sediment modeling and forecasting. This study aimed to develop and assess the applicability of MLP, CANFIS, and MLR models by training and testing them during the monsoon season (June to September) for the Hurdag watershed in the Damodar-Barakar basin, located in Hazaribagh district, Jharkhand, India. Daily rainfall, runoff (or streamflow), and sediment concentration data from 1997 to 2006 were used, with the data split into two sets: training set (1997–2004) and a testing set (2005–2006). The analysis was conducted using NeuroSolution 5.0 software and Microsoft Excel for performance evaluation indices. The best input combinations for sediment yield simulation were identified, and 10 optimal models were selected from 31 different input combinations. These input combinations were applied to the network for training using the back propagation algorithm for MLP and Gaussian and generalized bell membership functions for CANFIS models. Multiple networks were trained individually, with the most accurate predictions during testing being chosen as the best models. The models’ performance was evaluated using statistical indices such as root mean squared error (RMSE), coefficient of efficiency (CE), and correlation coefficient (r). The results showed that MLP and CANFIS models performed best in predicting sediment concentration for the Hurdag watershed, whereas MLR models showed poor performance for the given dataset. Specifically, sediment concentration for the current day could be modeled using current day rainfall and runoff data (MLP-7), while runoff could be simulated using the previous day's rainfall data (MLP-2).