Neural Network Topology Research Articles

Learning PDFs through interpretable latent representations in Mellin space

Representing the parton distribution functions (PDFs) of the proton and other hadrons through flexible, high-fidelity parametrizations has been a long-standing goal of particle physics phenomenology. This is particularly true since the chosen parametrization methodology can play an influential role in the ultimate PDF uncertainties as extracted in QCD global analyses; these, in turn, are often determinative of the reach of experiments at the LHC and other facilities to nonstandard physics, including at large x, where parametrization effects can be significant. In this study, we explore a series of encoder-decoder machine-learning (ML) models with various neural-network topologies as efficient means of reconstructing PDFs from meaningful information stored in an interpretable latent space. Given recent effort to pioneer synergies between QCD analyses and lattice-gauge calculations, we formulate a latent representation based on the behavior of PDFs in Mellin space, i.e., their integrated moments, and test the ability of various models to decode PDFs from this information faithfully. We introduce a numerical package, , which implements several encoder-decoder models to reconstruct PDFs with high fidelity and use this end-to-end tool to explore how such neural-network-based models might connect PDF parametrizations to underlying properties like their Mellin moments. We additionally dissect patterns of learned correlations between encoded Mellin moments and reconstructed PDFs that suggest opportunities for further improvements to ML-based approaches to PDF parametrizations and uncertainty quantification. Published by the American Physical Society 2025

Graph Neural Networks for Enhanced Social Network Analysis

Social network analysis (SNA) is an important approach for understanding complex linkages and interactions between entities. Traditional approaches frequently fail to capture the complexities of network data due to its non-Euclidean character. Graph Neural Networks (GNNs) offer an innovative approach to data analysis by modelling node, edge, and graph features using graph structures and neural network topologies. This study investigates the use of GNNs in social network analysis, focusing on problems such as community recognition, impact maximization, link prediction, and sentiment analysis. Our analysis of cutting-edge GNN models shows how they effectively capture and utilize topological and contextual information from social networks

Comparison of Some Optimizers in Long Short-Term Memory Networks with an Application to Tigris River Water Imports

The Long Short-Term Memory (LSTM) RNN architecture was developed to overcome the limitations of traditional RNNs in recognizing long-term dependencies in sequential input. LSTM networks and other neural network topologies rely heavily on optimizers for training. Every iteration of a gradient-based algorithm attempts to approach the minimizer/maximizer cost function by using the gradient's objective function information. Moreover, a comparative analysis of a number of optimizers is presented in this work, including FTRL, Nadam, Adagrad, Adadelta, SGD, RMSprop, and Adagrad. By directly affecting the objective function's convergence, the study emphasizes the crucial role optimizers play in improving gradient descent efficiency. This work fills a gap in the literature on choosing the best optimizers for time-series data by concentrating on water import statistics for the Tigris River. The results offer insightful information about how to choose optimizers wisely to reduce time complexity and increase accuracy. The results of the study show that these optimizers' performance is generally comparable when compared using the root mean square error (RMSE) criterion based on water data. However, Nadam produces RMSE values at 500 epochs that are lower than those of the other methods.

A density-based MS disease diagnosis model using the capuchin search algorithm and an ensemble of deep neural networks

Multiple sclerosis (MS) is a severe brain disease that permanently destroys brain cells, impacting vision, balance, muscle control, and daily activity. This research employs a weighted combination of deep neural networks and optimization techniques for MS disease diagnosis. This method uses slices of magnetic resonance imaging (MRI) images as input. Then, after the pre-processing operation, the process of segmentation and identification of the region of interest (ROI) is performed using a combination of the fuzzy c-means (FCM) algorithm and the capuchin search algorithm (CapSA) algorithm. When the target view is detected, the features of each ROI are extracted through three techniques: local binary pattern (LBP), multi-linear principal component analysis (MPCA), and gray level co-occurrence matrix (GLCM). Each of these features is then processed by a deep neural network. In each deep neural network, the CapSA algorithm is used to determine the optimal topology structure and adjust the weight vector of the neural network. This means that in this process, the vector and topology of the deep neural network are adjusted using the CapSA algorithm in such a way that the training error is minimized. Finally, after creating the trained models, the weighted combination of the outputs of these three models is used for the final diagnosis. The implementation results showed that our method was successful in achieving 100% precision compared to other comparative methods. Also, in the average accuracy criterion, it showed a performance of 99.51%, which shows the high performance of our method in diagnosing patients.

Event‐Triggered Synchronization Control for Markov Jump Neural Networks With Partially Unknown Transition Probabilities

ABSTRACTThis article studies the problem of static output feedback synchronization control of Markov jump neural networks. Given the randomness of the neural network topology and the limitations in acquiring transition probabilities, a Markov model with partially unknown transition probabilities is adopted, which aligns more closely with practical applications. To enhance communication efficiency in resource‐constrained environments, an event‐triggered mechanism is introduced. Additionally, in contrast to previous studies, this article employs the technique of free‐weighting matrix to address the decoupling issue in such neural networks, significantly reducing the conservativeness of the static output feedback control strategy. Finally, the theoretical findings are validated through simulation, demonstrating the practical applicability and effectiveness of the theoretical results.

Optimizing deep neural network architectures for renewable energy forecasting

An accurate renewable energy output forecast is essential for energy efficiency and power system stability. Long Short-Term Memory(LSTM), Bidirectional LSTM(BiLSTM), Gated Recurrent Unit(GRU), and Convolutional Neural Network-LSTM(CNN-LSTM) Deep Neural Network (DNN) topologies are tested for solar and wind power production forecasting in this study. ARIMA was compared to the models. This study offers a unique architecture for Deep Neural Networks (DNNs) that are specifically tailored for renewable energy forecasting, optimizing accuracy by advanced hyperparameter tuning and the incorporation of essential meteorological and temporal variables. The optimized LSTM model outperformed others, with MAE (0.08765), MSE (0.00876), RMSE (0.09363), MAPE (3.8765), and R2 (0.99234) values. The GRU, CNN-LSTM, and BiLSTM models predicted well. Meteorological and time-based factors enhanced model accuracy. The addition of sun and wind data improved its prediction. The results show that advanced deep neural network (DNN) models can predict renewable energy, highlighting the importance of carefully selecting characteristics and fine-tuning the model. This work improves renewable energy estimates to promote a more reliable and environmentally sustainable electricity system.

A generalized framework of neural networks for Hamiltonian systems

A generalized framework of neural networks for Hamiltonian systems

A modified neural network method for computing the Lyapunov exponent spectrum in the nonlinear analysis of dynamical systems

A modified neural network method for computing the Lyapunov exponent spectrum in the nonlinear analysis of dynamical systems

Research on the simulation method of a BP neural network PID control for stellar spectrum.

This study investigated the multiple correlations among spectral simulation units based on digital micromirror device (DMD) spectral simulation, which leads to the problem that conventional spectral simulation methods such as PID control exhibit a low fitting accuracy or long fitting time in the spectral simulation of various targets. In this paper, a method of stellar spectrum simulation based on back propagation neural network-based PID (BP-PID) control is proposed to achieve high efficiency and high precision simulation of various spectral targets. The topology of the BP neural network was constructed based on the spectral modulation model of a DMD stellar spectrum simulation system, and the algorithm of the BP-PID control was designed. Finally, an experimental platform was built to verify the performance and spectral simulation accuracy of the BP-PID control algorithm. The results show that the overshoot and response time of the BP-PID control algorithm decreased by 79.01% and 30%, respectively compared with those of the PID control algorithm. The maximum spectral simulation accuracies of 2000K, 7000K, and 12000K color temperature increased by a factor of 2.311, 1.871, and 2.254, respectively, and the standard deviations of the spectral simulation error decreased by 56%, 41%, and 54%, respectively. In the range of 2000-12000K color temperature, the spectral simulation error of the BP-PID control algorithm is better than ±3.495%, and the standard deviation of the spectral simulation error is between 1.8255 and 2.2358. The proposed method can improve the spectral simulation accuracy and simulation efficiency of a star simulator, reduce the magnitude and spectrum calibration errors caused by the differential response, improve the star feature recognition accuracy of the orbiting star sensor, and hence, provide a theoretical and technical basis for the development of high-precision star sensors.

How Can Anomalous-Diffusion Neural Networks Under Connectomics Generate Optimized Spatiotemporal Dynamics.

Spatiotemporal dynamics in the brain have been recognized as strongly related to the formation of perceived and cognitive diseases, such as delusions and hallucinations in Alzheimer's disease. However, two practical considerations are rarely mentioned in related mechanism research: the connectomics networking and the anomalous diffusion generated by the complex medium between neurons and the complex topology of neural networks, respectively. Furthermore, how to optimize the corresponding dynamics behaviors has excellent implications for treating brain diseases. This article first realizes the networking under connectomics for an anomalous-diffusion single-neuron model and applies a nonlinear state feedback control to generate optimized dynamic behaviors, which provides a paradigm of nonequilibrium self-organization driven by anomalous diffusion. Then, by tracing the root distribution of the characteristic equation, some controlled conditions causing or inhibiting Turing instability and Hopf bifurcation are deduced, and the effects of self-diffusion and cross diffusion on Turing instability range are also revealed. At last, thorough numerical simulations are updated to illustrate the results. It is emphasized that delay, self-diffusion, cross diffusion, and fractional order occupy dominant positions in determining the network's spatiotemporal dynamics, and utilizing the control strategy can efficiently reduce Turing instability and delay Hopf bifurcation.

Data-driven prediction of convective heat transfer coefficients in internal walls of aero-engine bearing chambers using Mind Evolution Algorithm-Enhanced Bayesian regularization neural networks

Data-driven prediction of convective heat transfer coefficients in internal walls of aero-engine bearing chambers using Mind Evolution Algorithm-Enhanced Bayesian regularization neural networks

Hub disruption in HIV disease and cocaine use: A connectomics analysis of brain function

Cocaine use (CU) is prevalent in people with HIV (PWH). Both conditions are linked to changes in cognitive functioning and neural network topology. The current study utilizes graph theory to investigate functional connectomics associated with HIV and CU, focusing on disruption of densely connected nodes called hubs. Resting state functional magnetic resonance imaging (fMRI) from 206 adults (ages 22-55 years) were analyzed. A HIV x CU factorial design was implemented with participants in four groups: HIV+CU (n= 41), HIV only (n= 88), CU only (n= 36), and controls (n= 41). Functional connectomes were constructed, and thresholded graph metrics were calculated. Network centrality metrics - betweenness centrality (BC), participation coefficient (PC), and within module degree (WD) - were quantified into hub disruption indices (HDI). For each index, a 2×2 ANCOVA was performed controlling for education. Participants were 68 % male and 74 % African-American with a mean age of 44.4 years. HIV and CU were associated with hub disruption in all three indices. Interactions were significant for HDI-PC and HDI-WD, such that HIV disease was associated with greater hub disruption among participants without CU, but not among participants with CU. Overall, lower global cognitive functioning was associated with greater hub disruption on all three indices. Widespread hub disruption was evident in HIV disease and CU, highlighting topological reorganization in both diseases with neurocognitive effects. Hub-related measures inform functional connectivity disruptions in HIV disease and CU, particularly with respect to changes in network topology throughout the connectome.

Optimizing the topology of convolutional neural network (CNN) and artificial neural network (ANN) for brain tumor diagnosis (BTD) through MRIs

The use of MRI analysis for BTD and tumor type detection has considerable importance within the domain of machine vision. Numerous methodologies have been proposed to address this issue, and significant progress has been achieved in this domain via the use of deep learning (DL) approaches. While the majority of offered approaches using artificial neural networks (ANNs) and deep neural networks (DNNs) demonstrate satisfactory performance in Bayesian Tree Descent (BTD), none of these research studies can ensure the optimality of the employed learning model structure. Put simply, there is room for improvement in the efficiency of these learning models in BTD. This research introduces a novel approach for optimizing the configuration of Convolutional Neural Networks (CNNs) and Artificial Neural Networks (ANNs) to address the BTD issue. The suggested approach employs Convolutional Neural Networks (CNN) for the purpose of segmenting brain MRIs. The model's configurable hyper-parameters are tuned using a genetic algorithm (GA). The Multi-Linear Principal Component Analysis (MPCA) is used to decrease the dimensionality of the segmented features in the pictures after they have been segmented. Ultimately, the segmentation procedure is executed using an Artificial Neural Network (ANN). In this artificial neural network (ANN), the genetic algorithm (GA) sets the ideal number of neurons in the hidden layer and the appropriate weight vector. The effectiveness of the suggested approach was assessed by utilizing the BRATS2014 and BTD20 databases. The results indicate that the proposed method can classify samples from these two databases with an average accuracy of 98.6% and 99.1%, respectively, which represents an accuracy improvement of at least 1.1% over the preceding methods.

Improving the performance of mutation-based evolving artificial neural networks with self-adaptive mutations.

Neuroevolution is a promising approach for designing artificial neural networks using an evolutionary algorithm. Unlike recent trending methods that rely on gradient-based algorithms, neuroevolution can simultaneously evolve the topology and weights of neural networks. In neuroevolution with topological evolution, handling crossover is challenging because of the competing conventions problem. Mutation-based evolving artificial neural network is an alternative topology and weights neuroevolution approach that omits crossover and uses only mutations for genetic variation. This study enhances the performance of mutation-based evolving artificial neural network in two ways. First, the mutation step size controlling the magnitude of the parameter perturbation is automatically adjusted by a self-adaptive mutation mechanism, enabling a balance between exploration and exploitation during the evolution process. Second, the structural mutation probabilities are automatically adjusted depending on the network size, preventing excessive expansion of the topology. The proposed methods are compared with conventional neuroevolution algorithms using locomotion tasks provided in the OpenAI Gym benchmarks. The results demonstrate that the proposed methods with the self-adaptive mutation mechanism can achieve better performance. In addition, the adjustment of structural mutation probabilities can mitigate topological bloat while maintaining performance.

Machine Learning in the Problem of Extrapolating Variational Calculations in Nuclear Physics

A modified machine learning method is proposed, utilizing an ensemble of artificial neural networks for the extrapolation of energies obtained in variational calculations, specifically in the No-core Shell Model (NCSM), to the case of the infinite basis. A new neural network topology is employed, and criteria for selecting both the data used for training and the trained neural networks for statistical analysis of the results are formulated. The approach is tested by extrapolating the deutron ground state energy in calculations with the Nijmegen II NN interaction and provides statistically significant results. This technique is applied to obtain extrapolated ground state energies of 6Li and 6He nuclei based on the NCSM calculations with Daejeon16 NN interaction.

Sparsity-Aware In-Memory Neuromorphic Computing Unit With Configurable Topology of Hybrid Spiking and Artificial Neural Network

Sparsity-Aware In-Memory Neuromorphic Computing Unit With Configurable Topology of Hybrid Spiking and Artificial Neural Network

Intelligent fault diagnosis of machinery based on hybrid deep learning with multi temporal correlation feature fusion

AbstractWith the advent of intelligent manufacturing era, higher requirements are put forward for the fault diagnosis technology of machinery. The existing data‐driven approaches either rely on specialized empirical knowledge for feature analysis, or adopt single deep neural network topology structure for automatic feature extraction with compromise of certain information loss especially the time‐series information's sacrifice, which both eventually affect the diagnosis accuracy. To address the issue, this paper proposes a novel multi‐temporal correlation feature fusion net (MTCFF‐Net) for intelligent fault diagnosis, which can capture and retain time‐series fault feature information from different dimensions. MTCFF‐Net contains four sub‐networks, which are long and short‐term memory (LSTM) sub‐network, Gramian angular summation field (GASF)‐GhostNet sub‐network and Markov transition field (MTF)‐GhostNet sub‐network and feature fusion sub‐network. Features of different dimensional are extracted through parallel LSTM sub‐network, GASF‐GhostNet sub‐network and MTF‐GhostNet sub‐network, and then fused by feature fusion sub‐network for accurate fault diagnosis. Two fault diagnosis experimental studies on bearings are implemented to validate the effectiveness and generalization of the proposed MTCFF‐Net. Experimental results demonstrate that the proposed model is superior to other comparative approaches.

Block-scale Oil Palm Yield Prediction Using Machine Learning Approaches Based on Landsat and MODIS Satellite Data

Due to environmental threats and weather uncertainty concerns, oil palm yield prediction is crucial for sustaining crop production. This can be achieved through machine learning and utilising remotely sensed data to predict crop yield. However, the comparative studies on remotely sensed data in adopting the machine learning models are still limited due to the data accessibility. Therefore, we compare and evaluate the prediction accuracy between different satellites, namely MODIS and Landsat-7, using machine learning algorithms and the topology of deep neural networks. Random forest and stacking outperformed linear regression, ridge regression, and lasso regression for both Landsat-7 NDVI (R2= 0.78–0.80; RMSE=1.00- 1.26 tonnes per hectare; MAE=0.77- 0.79 tonnes per hectares; MAPE=0.03-0.04 tonnes per hectare) and MODIS NDVI (R2= 0.60–0.65 tonnes per hectares; RMSE= 2.72–2.81 tonne per hectares; MAE= 1.42-1.55, MAPE= 1.01- 1.02 tonnes per hectares). The Landsat-7 NDVI revealed that neural networks with a deeper network topology (R2= 0.85; RMSE= 1.42 tonnes per hectare; MAE=0.57 tonnes per hectares; MAPE=0.06 tonnes per hectare) outperformed neural networks with a baseline and broader network topologies in terms of performance. In contrast, MODIS-NDVI revealed that the neural network with a wider network topology had the highest overall prediction accuracy and the lowest prediction error (R2= 0.75; RMSE= 2.81 tonnes per hectare; MAE=2.27 tonnes per tonnes; MAPE= 0.13). Because of its higher spatial resolution in comparison to MODIS, landsat-7 NDVI used in neural networks with a deep network topology provided the best model performance. Although the use of NDVI as a single input factor may cause uncertainty in some extents, it is an efficient and reliable method for improving yield estimation with the use of medium-resolution satellites, which has important implications for early warning towards the reduction in yield production.

WITHDRAWN: Application of Data Mining Technology in Analysis of Influencing Factors of Smart Electric Power Emergency Rescue

WITHDRAWN: Application of Data Mining Technology in Analysis of Influencing Factors of Smart Electric Power Emergency Rescue

Predictive process model monitoring using long short-term memory networks

Predictive process model monitoring using long short-term memory networks