Network Learning Research Articles

TinySpiking: a lightweight and efficient python framework for unsupervised learning spiking neural networks

Abstract Neural computation frameworks are essential for advancing computational neuroscience and artificial intelligence, offering a robust platform for simulating intricate brain-like processes and fostering the growth of intelligent systems. This study presents TinySpiking, a novel, lightweight, and energy-efficient Python framework designed for the simulation and learning of Spiking Neural Networks (SNNs). Unlike traditional frameworks, TinySpiking does not depend on third-party libraries, thereby reducing computational overhead and energy consumption. The framework's innovation is that it implements unsupervised learning through Spike-Time Dependent Plasticity (STDP), a biologically inspired learning rule that adjusts synaptic weights based on the precise timing of neuronal spikes. This mechanism is crucial for constructing complex neural architectures and effectively processing spatio-temporal data, which is vital for the dynamic and intricate demands of real-world data analysis. Demonstrating the practical utility of TinySpiking, we applied it to image reconstruction tasks using the MNIST and landmark datasets. The results not only validate the network's ability to autonomously identify and enhance key visual features without external supervision but also highlight its efficiency in learning from data, mirroring the adaptability of biological neural systems. In conclusion, TinySpiking's innovative, lightweight design and its proven effectiveness in unsupervised learning tasks make it a standout computational tool for the fields of computational neuroscience and machine learning. Its low time consumption, biological plausibility, and independence from third-party libraries position it as a compelling platform for future research and applications, promising to drive advancements in unsupervised learning and intelligent system development.

Neural mechanisms of relational learning and fast knowledge reassembly in plastic neural networks

Neural mechanisms of relational learning and fast knowledge reassembly in plastic neural networks

A Hot Rolling Full Process Rolling Force Prediction Method Based on Transfer Learning and Inception-LSTM Neural Network

A Hot Rolling Full Process Rolling Force Prediction Method Based on Transfer Learning and Inception-LSTM Neural Network

Speech Technology for Automatic Recognition and Assessment of Dysarthric Speech: An Overview.

In this review article, we present an extensive overview of recent developments in the area of dysarthric speech research. One of the key objectives of speech technology research is to improve the quality of life of its users, as evidenced by the focus of current research trends on creating inclusive conversational interfaces that cater to pathological speech, out of which dysarthric speech is an important example. Applications of speech technology research for dysarthric speech demand a clear understanding of the acoustics of dysarthric speech as well as of speech technologies, including machine learning and deep neural networks for speech processing. We review studies pertaining to speech technology and dysarthric speech. Specifically, we discuss dysarthric speech corpora, acoustic analysis, intelligibility assessment, and automatic speech recognition. We also delve into deep learning approaches for automatic assessment and recognition of dysarthric speech. Ethics committee or institutional review board did not apply to this study. Overcoming the challenge of limited data and exploring new avenues in data collection, artificial intelligence-powered analysis and teletherapy hold immense potential for significant advancements in dysarthria research. To make longer and faster strides, researchers typically rely on existing research and data on a global scale. Therefore, it is imperative to consolidate the existing research and present it in a form that can serve as a basis for future work. In this review article, we have reviewed the contributions of speech technologists to the area of dysarthric speech with a focus on acoustic analysis, speech features, and techniques used. By focusing on the existing research and future directions, researchers can develop more effective tools and interventions to improve communication, quality of life, and overall well-being for people with dysarthria.

W-Shaped Net: An Inter-Slice Super-Resolution Segmentation Deep Network for CT Scans of Hepatic Ducts

The three-dimensional (3D) reconstruction of the hepatic duct tree is significant for the minimally invasive surgery of hepatobiliary stone disease, which can be influenced by the quantity of the CT scans of hepatic ducts. If insufficient CT scans with low inter-slice resolution are directly utilized for 3D reconstruction, discontinuities and gaps will emerge in the reconstructed hepatic duct tree. In this paper, a novel end-to-end deep learning framework is designed for the inter-slice super-resolution segmentation of the CT slices of hepatic ducts, which can improve the 3D reconstruction performance in the inter-slice dimension. Specifically, the framework cascades into an inter-slice super-resolution subnetwork and a segmentation subnetwork. A deep learning network is introduced as the inter-slice super-resolution subnetwork to generate an intermediate slice between two adjacent CT slices in the simulated CT scans with low inter-slice resolution. To capture the spatiotemporal correlation existing in the CT scans of hepatic ducts, the ConvLSTM is introduced into the U-Net-like segmentation subnetwork in the high-dimensional feature space. To further suppress the problems of discontinuities and gaps, a structure-aware loss function is proposed by incorporating the structural similarity index measure (SSIM) as a regulator to dynamically assign the contribution of the generated CT slice to the total loss of the designed framework. Experimental results demonstrate that our proposed framework performs better segmentations for hepatic ducts than several existing deep learning networks with the performance of 0.7690 DICE and a 0.7712 F1-score, which is beneficial for the 3D reconstruction of the hepatic duct tree.

Rapid Identification of Saline–Alkali Stress-Tolerant Peanut Varieties Based on Multimodal Data

The cultivation of saline–alkali-tolerant peanut (Arachis hypogaea L.) varieties can effectively increase grain yield in saline–alkali land. However, traditional assessment methods are often cumbersome and time consuming. To rapidly identify saline–alkali stress-tolerant peanut varieties, this research proposed a saline–alkali stress tolerance evaluation method based on deep learning and multimodal data. Specifically, the research first established multimodal datasets for peanuts at different growth stages and constructed a saline–alkali stress score standard based on unsupervised learning. Subsequently, a deep learning network called BO-MFFNet was built and its structure and hyperparameters were optimized by the Bayes optimization (BO) algorithm. Finally, the point prediction of the saline–alkali stress score were carried out by using the Gaussian process regression model. The experimental results show that the multimodal method is superior to the single-modal data and the BO algorithm significantly improves the performance of the model. The root mean squared error and relative percentage deviation of the BO-MFFNet model are 0.089 and 3.669, respectively. The model effectively predicted the salt–alkali stress tolerance of five varieties, and the predicted results were Huayu25, Yuhua31, Yuhua33, Yuhua32, and Yuhua164 from high to low. This research provides a new method for assessing crop tolerance under extreme environmental stress.

Research on peanut pedigree analysis and variety identification methods based on deep features extraction and hierarchical clustering

Abstract In order to screen high-quality peanut pod varieties on food processing production lines and promote the sustainable development of the peanut as well as the expansion of its consumer market, this study improved the recognition ability of peanut pod varieties and the efficiency of deep feature extraction by optimizing the ResNet50 deep learning network model. Experimental results showed that the accuracy of the optimized network model reached 91.6%, which was 2.1% higher than the original ResNet50 model. Additionally, this study extracted the deep features of the appearance morphology of peanut pods, used the agglomerative clustering method to explore the genetic relationship of hybrid offspring varieties under the same line, and constructed a pedigree diagram containing 18 varieties. These research findings provide a crucial scientific foundation for the cultivation of high-quality peanut varieties. The selected high-quality peanuts not only enhance the added value of the peanut industry but also further expand the market potential of peanut by-products.

Optimising torque for three-disc afpmsm in electric vehicles using bp_ann and anfis algorithms

Designing an optimal torque distribution controller for the three-disc axial flux permanent magnet synchronous (three-disc AFPMSM) optimises performance while ensuring robustness, stability, and adaptability in a real-world condition. This is crucial for maximising the potential of AFPMSM, particularly in a modern application like an electric vehicle and a renewable energy. Thus, a controller compatible with more complex systems in the future is essential. This paper presents a system that combines torque control algorithms based on a back-propagation neural network (BP-ANN) and an Adaptive Neuro-Fuzzy Inference System (ANFIS). The BP-ANN uses a multi-layer structure where the input layer processes factors such as load torque, rotational speed, and stator current, hidden layers model complex nonlinear interactions, and the output layer predicts optimal torque for AFPMSM operation. Training involves minimising the error between predicted and actual torque through gradient descent and iterative adjustments of weights and biases. The ANFIS-based control enhances performance by integrating neural network learning with fuzzy logic to optimise torque output. By leveraging the strengths of both BP-ANN and ANFIS, the system offers a stable, efficient, and adaptable solution for three-disc AFPMSMs. The Matlab/Simulink simulations confirm its effectiveness, showing balanced torque distribution, reduced energy losses, improved drivetrain efficiency, and adaptability to sudden load or road changes, ensuring stability and enhanced dynamic response

A Novel CNN-Based Framework for Alzheimer’s Disease Detection Using EEG Spectrogram Representations

Background: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that poses critical challenges in global healthcare due to its increasing prevalence and severity. Diagnosing AD and other dementias, such as frontotemporal dementia (FTD), is slow and resource-intensive, underscoring the need for automated approaches. Methods: To address this gap, this study proposes a novel deep learning methodology for EEG classification of AD, FTD, and control (CN) signals. The approach incorporates advanced preprocessing techniques and CNN classification of FFT-based spectrograms and is evaluated using the leave-N-subjects-out validation, ensuring robust cross-subject generalizability. Results: The results indicate that the proposed methodology outperforms state-of-the-art machine learning and EEG-specific neural network models, achieving an accuracy of 79.45% for AD/CN classification and 80.69% for AD+FTD/CN classification. Conclusions: These results highlight the potential of EEG-based deep learning models for early dementia screening, enabling more efficient, scalable, and accessible diagnostic tools.

Deformation Detection Method for Substation Noise Barrier Column Based on Deep Learning and Digital Image Technology

The dynamic identification of the deformation of a noise barrier column is of great significance to the monitoring of its health. At the same time, the maximum stress of the column is an important indicator for the evaluation of its health status. Traditional contact displacement monitoring installs sensors on the structure, requires a lot of wiring and data acquisition equipment, and establishes a relatively independent and stable displacement reference system. Affected by the environment, wear, and material aging, the efficiency and reliability of data acquisition are reduced. A monitoring method based on digital image has the advantages of non-contact monitoring, high precision, and strong reliability. The existing DIC detection methods are limited by processor performance and image resolution, which are difficult to apply to engineering detection. In this paper, a structural displacement identification method based on convolutional neural networks (CNNs) and DIC technology is proposed. In this method, the data set is formed according to the column displacement cloud image obtained by DIC analysis, and the data set is enhanced by data normalization and region division. Through the analysis of the number of network layers and learning rate, the model design of the deep learning network is carried out. The high-speed camera image results of the test are introduced and identified by the static loading test of the equal-scale sound barrier. The results show that the structural displacement identification method based on CNN and DIC technology can accurately identify the displacement change in the structure, which greatly improves the efficiency of image displacement calculation using DIC technology.

Multivariate time series weather forecasting model using integrated secondary decomposition and Self-Attentive spatio-temporal learning network

<p>Weather forecasting is an essential but complex task because of its substantial influence on human life and advanced atmospheric dynamics. In recent years, deep learning-based techniques have gain high attention for weather forecasting. Nevertheless, current forecasting models ignore the interdependent relationship between variables across regions and primarily examine temporal patterns of weather data. In this work, a new Multivariate time series weather forecasting model is proposed using integrated secondary decomposition and Self-Attentive spatio-temporal learning network (SASTLNet). Particularly, the weather data is filtered using a Singular Spectrum Decomposition with Fuzzy Entropy (SSD-FE) for removing the irrelevant components and screening the component containing vital information. Moreover, the spatiotemporal fluctuations in the meteorological variables are investigated using an enhanced empirical mode decomposition (E3MD) method and an SASTLNet model. The self-attention block of the SASTLNet model is discovered using ladder format to lower the processing costs while representing the temporal and spatial features through single memory cell. The simulation results prove that the suggested model can outperforms the baseline models in terms of efficacy and accuracy.</p>

SSSA: low data sentiment analysis using boosting semi-supervised approach and deep feature learning network

SSSA: low data sentiment analysis using boosting semi-supervised approach and deep feature learning network

A Deep Multi-Task Learning Model for OD Traffic Flow Prediction Between Highway Stations

The rapid development of highways greatly affects the flow of people, finance, goods, and information between cities, and monitoring the OD flow of travel has become a very important task for intelligent transportation systems (ITS). The temporal dynamics and complex spatial correlations of OD traffic distribution, as well as the sparsity and incompleteness of data caused by uneven traffic distribution, make OD traffic prediction complex and challenging. This paper proposes a multi-task prediction model for OD traffic between highway stations. The model adopts a hard parameter shared multi-task learning network structure, which is divided into sub-task learning inflow trend modules, sub-task learning outflow trend modules, and main task learning modules for OD traffic. At the same time, the attraction intensity matrix between stations is constructed using the population density data as the external feature of the sub-task module for outlet outflow flow, and stronger constraints between tasks are introduced to achieve better fitting results. Finally, an OD flow prediction case experiment was conducted between stations on highways in Sichuan Province. The experimental results showed that the proposed model not only had higher accuracy in predicting results than other baseline models, but also had better effectiveness and robustness.

Proactive Resilient Adaptive Sensor Node Management with Neuro-Symbolic Systems for Optimized Coverage Efficiency in the Mission-Critical Applications

Abstract Wireless sensor networks (WSNs) play a critical role in applications such as wildlife monitoring, disaster recovery, and precision agriculture, where continuous coverage and longevity are paramount amidst dynamic environmental challenges. To address these demands, the cellular adaptive energy forecasting and coverage optimization (CAEFCO) framework integrates localized neuro-symbolic energy forecasting (LNS-EF), a novel concept that combines symbolic reasoning with neural network learning directly on sensor nodes. LNS-EF enables nodes to not only predict energy depletion based on past consumption patterns and environmental factors but also incorporate rule-based contextual reasoning for enhanced decision-making. Alongside this, CAEFCO employs an anomaly detection module that identifies disruptions, such as sensor damage or environmental interference, allowing real-time task redistribution. This dual approach ensures seamless task reallocation while extending network lifetime. CAEFCO’s proactive methodology demonstrates a 97% reduction in data loss and an 85% improvement in network longevity, offering a breakthrough in the resilience and sustainability of WSNs in mission-critical and harsh environments.

A Study on the Differences in Optimized Inputs of Various Data-Driven Methods for Battery Capacity Prediction

As lithium-ion batteries become increasingly popular worldwide, accurately determining their capacity is crucial for various devices that rely on them. Numerous data-driven methods have been applied to evaluate battery-related parameters. In the application of these methods, input features play a critical role. Most researchers often use the same input features to compare the performance of various neural network models. However, because most models are regarded as black-box models, different methods may show different dependencies on specific features given the inherent differences in their internal structures. And the corresponding optimal inputs of different neural network models should be different. Therefore, comparing the differences in optimized input features for different neural networks is essential. This paper extracts 11 types of lithium battery-related health features, and experiments are conducted on two traditional machine learning networks and three advanced deep learning networks in three aspects of input differences. The experiment aims to systematically evaluate how changes in health feature types, dimensions, and data volume affect the performance of different methods and find the optimal input for each method. The results demonstrate that each network has its own optimal input in the aspects of health feature types, dimensions, and data volume. Moreover, under the premise of obtaining more accurate prediction accuracy, different networks have different requirements for input data. Therefore, in the process of using different types of neural networks for battery capacity prediction, it is very important to determine the type, dimension, and number of input health features according to the structure, category, and actual application requirements of the network. Different inputs will lead to larger differences in results. The optimization degree of mean absolute error (MAE) can be improved by 10–50%, and other indicators can also be optimized to varying degrees. Therefore, it is very important to optimize the network in a targeted manner.

The Role of Learning Analytics in Evaluating Course Effectiveness

This study aims to examine the use of learning analytics in course evaluation within higher education institutions, in order to identify effective methodologies and best practices for leveraging data to improve educational effectiveness. Following the PRISMA guidelines, a systematic literature search was conducted in Scopus, yielding 34 relevant studies published between 2015 and 2024 for analysis. The results reveal six key categories of learning analytics applications: sentiment analysis, questionnaire analysis, engagement analysis, topic classification, predictive modelling, and performance analysis. The data sources for learning analytics applications primarily include questionnaires and learning management systems. While descriptive analysis was found to be the most commonly employed analytical technique, advanced techniques such as machine learning, artificial intelligence, and social network analysis are becoming more prominent. The studies addressed a wide range of elements associated with course evaluation, including course design, content quality, assignments, instructional strategies, workload, feedback mechanisms, and the integration of technology. These findings highlight the importance of adopting holistic approaches to capture the multifaceted nature of student experiences. This study also uncovers major limitations in the existing research, such as small sample sizes, potential biases due to the use of survey-based methods, and challenges in generalising findings across disciplines. These insights underscore the need for further research to enhance the methodologies used in course evaluations. This study contributes to advancing learning analytics practices and emphasises the importance of innovative approaches for evaluating and improving course effectiveness.

Functional brain activations correlated with association strength and prediction error during novel symbol–speech sound learning

Abstract Efficient learning of letters–speech sound associations results in the specialization of visual and audiovisual brain regions, which is crucial for the development of proficient reading skills. However, the brain dynamics underlying this learning process remain poorly understood, and the involvement of learning and performance monitoring networks remains underexplored. Here we applied two mutually dependent feedback learning tasks in which novel symbol–speech sound associations were learned by 39 healthy adults. We employed functional magnetic resonance (fMRI) along with a reinforcement learning drift diffusion model to characterize trial-by-trial learning in behavior and brain. The model-based analysis showed that posterior–occipital activations during stimulus processing were positively modulated by trial-wise learning, as indicated by the increase in association strength between audiovisual pairs. Prediction errors, describing the update mechanism to learn from feedback across trials, modulated activations in several mid-frontal, striatal, and cingulate regions. Both tasks yielded similar patterns of results, despite differences in their relative difficulty. This study elucidates the processes involved in audiovisual learning that contribute to rapid visual specialization within a single experimental session and delineates a set of coactivated regions involved in learning from feedback. Our paradigm provides a framework to advance our understanding of the neurobiology of learning and reading development.

Network Pharmacology Approach and Experimental Verification to Explore the Anti-NSCLC Mechanism of Grifolic Acid

Lung cancer is the leading cause of cancer-related death. Non–small cell lung cancer (NSCLC) accounts for 85% of all lung cancers and over 60% express wild-type EGFR (WT-EGFR); however, EGFR tyrosine kinase inhibitors (TKIs) have limited effect in most patients with WT-EGFR tumors. In this study, we applied network pharmacology screening and MTT screening of bioactive compounds to obtain one novel grifolic acid that may inhibit NSCLC through the EGFR-ERK1/2 pathway. Through the PPI network and machine learning, we identified two hub genes, EGFR and AKT1, as potential therapeutic targets. Molecular docking confirmed that the grifolic acid could effectively bind to the key target, EGFR. Using the NSCLC cell line NCI-H1781 as an in vitro model, we evaluated the effect of the drugs’ combination on viability, apoptosis, and clonogenicity capacity. In vitro studies showed that combined treatment decreased cell viability, increased activation PARP, and caused cell cycle redistribution and significantly greater inhibition of pEGFR and pAKT. This study not only provides new insights into the mechanism of grifolic acid against NSCLC but also important information and new research ideas for the discovery of anti-NSCLC compounds from natural products.

Identification of key genes associated with oxidative stress in ischemic stroke via bioinformatics integrated analysis

BackgroundIschemic stroke (IS) is a common cerebrovascular disease. Although the formation of atherosclerosis, which is closely related to oxidative stress (OS), is associated with stroke-related deaths. However, the role of OS in IS is unknown.MethodsOS-related key genes were obtianed by overlapping the differentially expressed genes (DEGs) between IS and normal control (NC) specimens, IS-related genes, and OS-related genes. Then, we investigated the mechanism of action of key genes. Subsequently, protein–protein interaction (PPI) network and machine learning algorithms were utilized to excavate feature genes. In addition, the network between feature genes and microRNAs (miRNAs) was established to investigate the regulatory mechanism of feature genes. Finally, quantitative PCR (qPCR) was utilized to validate the expression of feature genes with blood specimens.ResultsA total of 42 key genes related to OS were acquired. Enrichment analysis indicated that the key genes were associated with oxidative stress, reactive oxygen species, lipid and atherosclerosis, and cell migration-related pathways. Then, 6 feature genes (HSPA8, NCF2, FOS, KLF4, THBS1, and HSPA1A) related to OS were identified for IS. Besides, 6 feature genes and 255 miRNAs were utilized to establish a feature genes-miRNA network which contained 261 nodes and 277 edges. At last, qPCR results revealed that there was a trend for higher expression of FOS, KLF4, and HSPA1A in IS specimens than in NC specimens. Additionally, HSPA8 expression was significantly decreased in the IS specimens, which was consistent with the findings of the GEO database analysis.ConclusionIn conclusion, 6 feature genes (HSPA8, NCF2, FOS, KLF4, THBS1, and HSPA1A) related to OS were mined by bioinformatics analysis, which might provide a new insights into the evaluation and treatment of IS.Clinical trial number: Not applicable.

Proactive Resilient Adaptive Sensor Node Management with Neuro-Symbolic Systems for Optimized Coverage Efficiency in the Mission-Critical Applications

Abstract Wireless sensor networks (WSNs) play a critical role in applications such as wildlife monitoring, disaster recovery, and precision agriculture, where continuous coverage and longevity are paramount amidst dynamic environmental challenges. To address these demands, the cellular adaptive energy forecasting and coverage optimization (CAEFCO) framework integrates localized neuro-symbolic energy forecasting (LNS-EF), a novel concept that combines symbolic reasoning with neural network learning directly on sensor nodes. LNS-EF enables nodes to not only predict energy depletion based on past consumption patterns and environmental factors but also incorporate rule-based contextual reasoning for enhanced decision-making. Alongside this, CAEFCO employs an anomaly detection module that identifies disruptions, such as sensor damage or environmental interference, allowing real-time task redistribution. This dual approach ensures seamless task reallocation while extending network lifetime. CAEFCO’s proactive methodology demonstrates a 97% reduction in data loss and an 85% improvement in network longevity, offering a breakthrough in the resilience and sustainability of WSNs in mission-critical and harsh environments.