Input To Network Research Articles

Support of Accounting by Bank Statement Classification Using Neural Networks

ABSTRACT With the advent of the era of Big Data, machine learning techniques are widely used in many technologies and industries. In machine learning, neural networks are used to recognize relationships between data. In this paper, neural networks are developed to support accounting processes. The neural networks use the texts from the bank statements as input for the neural network to predict the debit and credit account as well as the value-added tax percentage on the account bank statement. Fifteen different neural network models were built and evaluated based on accuracy and loss. The best-found neural network model is described in detail. The paper also discusses issues of data quality and pre-processing techniques. JEL Classifications: C45; D83; M41.

An Internet of Things Task Scheduling Framework Based on Agile Virtual Network on Demand Service Model

In order to improve the efficiency of cloud computing resource utilization and avoid the problem of computing resource allocation and scheduling lagging behind load changes, the author proposes a cloud computing resource on-demand allocation and elastic scheduling method based on network load prediction. Firstly, the author takes the network load data of Wikimedia as the research object and proposes an adaptive two-stage multi network model load prediction method based on LSTM, i.e. ATSMNN-LSTM load prediction method). This method can classify the network load data into climbing and descending types based on the trend and characteristics of the input network load data, And adaptively schedule the input network load data to the LSTM load prediction model that matches its type for prediction based on the classification results. The author proposes a maximum cloud service revenue computing resource quantity search algorithm based on network load prediction (i.e. MaxCSPRNWP algorithm), which aims to improve cloud service revenue as the optimization objective. Under the premise of ensuring task service quality and system stability, the algorithm allocates cloud computing resources on demand and flexibly schedules them in advance based on the predicted network load results. The experimental results show that the ATSMNNLSTM load prediction method proposed by the author can obtain more accurate network load prediction results compared to other load prediction methods, and the MaxCSPR-NWP algorithm, which is based on network load prediction and is capable of effectively converting the network load prediction results into the required number of cloud servers, is the maximum cloud service revenue computing resource quantity search algorithm proposed by the author, not only does it achieve the early allocation and scheduling of cloud computing resources, thereby avoiding the impact of lagging behind in computing resource allocation and scheduling due to load changes on the quality of cloud computing task services and resource utilization efficiency, at the same time, it has also achieved on-demand allocation and flexible scheduling of cloud computing resources with the goal of improving cloud service revenue.

Thermophysical properties, and figures-of-merit analyses of nanodiamond/CuO ionanofluids: Experimental and artificial neural network predictions

The nanodiamond/copper oxide (ND/CuO) nanoparticles were dispersed into 1-ethyl-3-methylimidazolium chloride [EMiM]Cl to create ionanofluids. The experiments were performed to analyse the thermophysical properties under particle volume loadings from 0.2 % to 1.0 %, and temperature ranging from 30 to 60 °C. Theoretical approach was made to understand the Figures-of-Merit (FoM) of the prepared ionanofluids, while they flow in a tube, and helical coil through the evaluated thermophysical properties. The lower thermal conductivity augment of 14.27 % was noticed at 0.2% vol. and larger thermal conductivity enhancement of 35.20 % was noticed at 1.0% vol. and at a temperature of 60 °C, against the base fluid. Moreover, the viscosity of 1.0% vol. of ionanofluid was enhanced by 61.66 % at a temperature of 30 °C, over the base fluid. The other properties like density was increased with an increase of volume loadings, and specific heat was decreased with an increase of volume loadings. The FoM results indicated that, all the prepared volume concentrations of ionanofluids flow in a tube was benefited fluid when the fluid operating temperature was exceeds to 35 °C, on the other side, the prepared ionanofluids flow in a helical coil was benefited fluid when its operating temperature was above 30 °C for 0.8 %, and 1.0% vol. loadings. The estimated thermophysical properties were also used as input and output data for an artificial neural network of Levenberg-Marquardt (LM) algorithm. It is noticed from the LM algorithm results, the predicted data is well agreed with an experimental data. For all the data, the correlation coefficient (R2) of thermal conductivity is 0.99283, viscosity is 0.9971, density is 0.99887, and specific heat is 0.99817, was observed, respectively.

Real-time prediction of metacentric height of ro-ro passenger ships in Qiongzhou strait based on improved RBF neural network

Addressing the complexities and real-time challenges in calculating ship metacentric height (GM), this study proposes an improved method using an optimized radial basis function neural network (RBFNN) for real-time GM prediction. Bayesian optimization is introduced to fine-tune the hyperparameters of the RBFNN, aiming to enhance the model's generalization performance. The study focuses on the Qiongzhou Strait Ro-ro passenger ship ‘Zijing No.11’ and selects three GM-related factors as neural network inputs using grey correlation analysis. The GM calculated by the empirical formula serves as the expected value, which is compared with predictions from various algorithms. Simulation results demonstrate that the improved RBFNN achieves significantly lower prediction errors compared to the unoptimized version. Furthermore, compared to other machine learning models and artificial neural networks, the proposed model exhibits superior performance in predicting ship initial stability height. Consequently, this model offers a practical tool for accurate and real-time GM prediction, enhancing intelligent stowage and operational efficiency in shipping.

A deep learning approach for pose error prediction in parallel robots

Parallel robots are increasingly favored in industrial automation due to their high precision, stability, and rigidity in high-speed and heavy-load tasks. However, factors such as manufacturing tolerances, assembly deviations, loads, and environmental conditions can lead to pose inaccuracies, manifesting as both geometric and non-geometric errors. Traditional model-driven approaches effectively predict geometric errors but often neglect non-geometric factors, potentially impacting the overall accuracy. This paper presents a deep learning-based method for comprehensive pose error prediction in parallel robots, addressing both geometric and non-geometric errors. The proposed approach employs an encoder-decoder network that takes target pose inputs and outputs predicted pose errors. The encoder integrates multi-layer convolutional sparse coding (CSC) blocks and efficient channel attention (ECA) modules to extract features from the network inputs. Monte Carlo dropout is utilized to estimate prediction uncertainties, enhancing the reliability of the network predictions. Experimental validation on a Stewart platform demonstrates the high accuracy and robustness of the method under various loading conditions. Comparative analyses with several typical models affirm the superior performance of the proposed network, indicating its potential to improve the operational accuracy and reliability of parallel robots in high-precision applications.

A novel wind turbine fault diagnosis method based on compressive sensing and lightweight SqueezeNet model

The fault diagnosis method based on model and signal processing has some problems, such as difficulty in modeling and difficulty in extracting signal features; As the depth of the neural network deepens, there will be the problem of gradient vanishing or mass, and directly converting the fault signal into a one-dimensional or two-dimensional image as the network input cannot retain the time information between the signals, resulting in the loss of information. In order to solve the above problems, this paper proposes a fault diagnosis method based on Compressed Sensing (CS) and lightweight SqueezeNet model. Firstly, the Compressive Sensing (CS) technology was used to sparse the original signal, and the Compressed Sampling Matching Pursuit (CoSaMP) algorithm was used to compress and reconstruct the signal to complete the signal noise reduction and remove the data redundancy. Secondly, the compressed and reconstructed signal was encoded by recursive graph (RP) to generate a two-dimensional image, and the temporal correlation characteristics of the signal were retained. Thirdly, the traditional SqueezeNet model is improved by using the residual idea to enhance the feature extraction ability. The RP dataset was input into the improved SqueezeNet model for fault feature extraction and fault classification. Finally, the experimental results show that compared with other methods, the proposed method can identify faults and have high diagnostic accuracy on the rolling bearing data samples of wind turbines.

Use of deep learning to improve seismic data quality analysis

Seismic data are influenced by various types of noise, which are typically categorized into two primary classes: anthropogenic and environmental. However, the detection of instrumental noise or malfunctioning stations also plays a crucial role in ensuring the data quality and the efficiency of a seismic network. The visual inspection of seismic spectral diagrams (e.g. power spectral density) enables us to identify issues that could potentially compromise data quality, thereby affecting subsequent calculations such as Magnitude or Peak Ground Acceleration (PGA). However, this process is time-consuming and demands significant human expertise due to the complexity of the diagrams, compounded by the sheer number of stations requiring examination. Therefore, in this paper, we explore the feasibility of transferring human expertise into an artificial intelligence system to create an automated system capable of rapidly performing such detection. More specifically, in the first part of this paper, we use Probability Density Function (PDF) diagrams, enabling an initial assessment of station performance via visual inspection. We describe this plot type and provide examples that reveal whether a station is functioning correctly or if technical issues exist. A table containing the main evaluation criteria is provided. In the second part of this paper, we demonstrate that these plots can serve as input for a neural network, allowing the development of the aforementioned automated system. Through extensive testing under various conditions, we have observed that the trained network consistently achieves an accuracy rate exceeding 85% across all four conductedtests. In the latest and most significant test, the achieved accuracy is approximately 87%.

Multi-part kinematic constraint prediction for automatic generation of CAD model assemblies using graph convolutional networks

This paper presents a machine learning-based approach to predict kinematic constraints between CAD models that have potentially never been assembled together before. During the learning phase, the algorithm is trained to predict the next-possible-constraints between a set of parts candidate to the assembly. Assemblies are represented in a new graph-based formalism that is capable of capturing features associated with parts, interfaces between parts and constraints between them. Using such a multi-level feature extraction strategy coupled to a state-by-state graph decomposition, the approach does not need to be trained on a large database. This formalism is used to model both the network input and output where the next-possible-constraints appear after evaluation. The core of the approach relies on a series of networks based on a link-prediction encoder–decoder architecture, integrating the capabilities of several convolutional networks trained in an end-to-end manner. A decision-making algorithm is added to post-process the output and drive the prediction process in finding one among the set of next-possible-constraints. This process is repeated until no more constraints can be added. The experimental results show that the proposed approach outperforms state-of-the-art methods on such assembly tasks. Although the state-by-state assembly algorithm is iterative, it still takes into account the whole set of parts as well as the whole set of constraints already predicted, and this makes it possible to handle constraint cycles, which is generally not possible when not considering multiple parts as input.

SURVEY OF ADVERSARIAL ATTACKS AND DEFENSE AGAINST ADVERSARIAL ATTACKS

In recent years, the fields of Artificial Intelligence (AI) and Deep learning (DL) techniques along with Neural Networks (NNs) have shown great progress and scope for future research. Along with all the developments comes the threats and security vulnerabilities to Neural Networks and AI models. A few fabricated inputs/samples can lead to deviations in the results of the models. Patch based Adversarial Attacks can change the output of a neural network to a completely different result just by making a few changes to the input of the neural network. These attacks employ a patch that is applied to the input image in order to cause the classifier to misclassify and make the incorrect prediction. The goal of this research is to develop effective defense strategies against these types of attacks and make the model/Neural Network more robust.

Identification of the damage location for the structural sealant based on deep learning

Structural sealants are essential to maintain the safety of panel units of hidden frame glass curtain wall. Damage localization of the concealed structural sealant is still difficult because of the unknown baseline model. In this paper, a convolutional neural network-based identification method is proposed to localize the damage of structural sealants without the baseline model. The method develops a novel input of the convolutional neural network (CNN), multi-symmetry-point images (MSPI), which is encoded by the response of four symmetrical points to impulse excitations. The CNN would identify the damage location by discerning differences among four images. Then, a dataset with 28803 samples, which considers the effect of the multiple damages and noise, was used to train the CNN. Three types of transformed images and four CNN models were compared to optimize the input signals and the configuration of the CNN. A series of numerical examples indicated that the Gram angle difference field is the optimal image transformation method, and improved-DenseNet121 is the optimal CNN for damage localization in structural sealants. Then, several laboratory experiments validate the effectiveness of the optimized image input and CNN with an accuracy rate of 92 % in the identification of damage location.

Photovoltaic Power Prediction Based on NRS-PCC Feature Selection and Multi-Scale CNN-LSTM Network

To improve the quality of photovoltaic (PV) data and power prediction accuracy, a PV power prediction method based on neighborhood rough set and Pearson correlation coefficient (NRS-PCC) feature selection and multi-scale convolutional neural networks and long short-term memory (CNN-LSTM) network is proposed. We first calculate the correlation between different PV features based on PCC and select strongly correlated features to cross-multiply to get the fusion features to enrich the data source. Then, dimensionality reduction of the fusion features by NRS. Finally, correlation analysis based on PCC on the dimensionality reduction of the fusion features to screen out the effective features. Furthermore, a multi-scale CNN-LSTM is designed to predict PV power. The output vectors of different convolutional layers are first fused to extract multi-scale features, then the features of different scales are spliced as the input of the LSTM network, and finally, the LSTM network performs regression prediction. The effectiveness of the proposed method is verified on a real PV power dataset.

Deep beamforming for speech enhancement and speaker localization with an array response-aware loss function

Recent research advances in deep neural network (DNN)-based beamformers have shown great promise for speech enhancement under adverse acoustic conditions. Different network architectures and input features have been explored in estimating beamforming weights. In this paper, we propose a deep beamformer based on an efficient convolutional recurrent network (CRN) trained with a novel ARray RespOnse-aWare (ARROW) loss function. The ARROW loss exploits the array responses of the target and interferer by using the ground truth relative transfer functions (RTFs). The DNN-based beamforming system, trained with ARROW loss through supervised learning, is able to perform speech enhancement and speaker localization jointly. Experimental results have shown that the proposed deep beamformer, trained with the linearly weighted scale-invariant source-to-noise ratio (SI-SNR) and ARROW loss functions, achieves superior performance in speech enhancement and speaker localization compared to two baselines.

Spectrum Sensing Method Based on STFT-RADN in Cognitive Radio Networks.

To address the common issues in traditional convolutional neural network (CNN)-based spectrum sensing algorithms in cognitive radio networks (CRNs), including inadequate signal feature representation, inefficient utilization of feature map information, and limited feature extraction capabilities due to shallow network structures, this paper proposes a spectrum sensing algorithm based on a short-time Fourier transform (STFT) and residual attention dense network (RADN). Specifically, the RADN model improves the basic residual block and introduces the convolutional block attention module (CBAM), combining residual connections and dense connections to form a powerful deep feature extraction structure known as residual in dense (RID). This significantly enhances the network's feature extraction capabilities. By performing STFT on the received signals and normalizing them, the signals are converted into time-frequency spectrograms as network inputs, better capturing signal features. The RADN is trained to extract abstract features from the time-frequency images, and the trained RADN serves as the final classifier for spectrum sensing. Experimental results demonstrate that the STFT-RADN spectrum sensing method significantly improves performance under low signal-to-noise ratio (SNR) conditions compared to traditional deep-learning-based methods. This method not only adapts to various modulation schemes but also exhibits high detection probability and strong robustness.

Multi-objective QoS optimization in swarm robotics

The “Internet of Robotic Things” (IoRT) is a concept that connects sensors and robotic objects. One of the practical applications of IoRT is swarm robotics, where multiple robots collaborate in a shared workspace to accomplish assigned tasks that may be challenging or impossible for a single robot to conquer. Swarm robots are particularly useful in critical situations, such as post-earthquake scenarios, where they can locate survivors and provide assistance in areas inaccessible to humans. In these life-saving situations, reliable and prompt communication among swarm robots is of utmost importance. To address the need for highly dependable and low-latency communication in swarm robotics, this research introduces a novel hybrid approach called Multi-objective QoS optimization based on Support vector regression and Genetic algorithm (MQSG). The MQSG method consists of two main phases: Parameter Relationship Identification and Parameter Optimization. In the Parameter Relationship Identification phase, the relationship between network inputs (Packet inter-arrival time, Packet size, Transmission power, Distance between sender and receiver) and outputs (quality of service (QoS) parameters) is established using support vector regression. In the parameter optimization phase, a multi-objective function is created based on the obtained relationships from the Parameter Relationship Identification phase. By solving this multi-objective function, optimal values for each QoS parameter are determined, leading to enhanced network performance. Simulation results demonstrate that the MQSG method outperforms other similar algorithms in terms of transmission latency, packet delivery rate, and the number of retransmitted packets.

Sink location problems in dynamic flow grid networks

A dynamic flow network consists of a directed graph, where nodes called sources represent locations of evacuees, and nodes called sinks represent locations of evacuation facilities. Each source and each sink are given supply representing the number of evacuees and demand representing the maximum number of acceptable evacuees, respectively. Each edge is given capacity and transit time. Here, the capacity of an edge bounds the rate at which evacuees can enter the edge per unit time, and the transit time represents the time which evacuees take to travel across the edge. The evacuation completion time is the minimum time at which each evacuee can arrive at one of the evacuation facilities. Given a dynamic flow network without sinks, once sinks are located on some nodes or edges, the evacuation completion time for this sink location is determined. We then consider the problem of locating sinks to minimize the evacuation completion time, called the sink location problem. The problems have been given polynomial-time algorithms only for limited networks such as paths [1–3], cycles [1], and trees [4–6], but no polynomial-time algorithms are known for more complex network classes. In this paper, we prove that the 1-sink location problem can be solved in polynomial-time when an input network is a grid with uniform edge capacity and transit time.

Centrality of Information Sources in Agricultural Input Networks: Insights from Social Network Analysis

Input dealers being credible localite sources of information for farmers have huge potential of aiding faster diffusion of innovations. Under this context, understanding how the input dealers’ access and exchange information is essential for enriching them with updated information and improving the agricultural value chain. This study employs Social Network Analysis (SNA) to explore the centrality measures—degree centrality, closeness centrality, and betweenness centrality—of various information sources utilized by input dealers in Nagar Kurnool and Wanaparthy districts of Telangana during 2020. The analysis reveals that traditional sources, such as extension agents and other input dealers, are highly central to the network, with both scoring the highest in degree and closeness centrality (0.733 and 0.698, respectively). These sources are not only frequently accessed but also serve as key bridges within the network, facilitating the flow of information. Digital platforms like smartphones show growing importance, with significant degree centrality (0.667), yet they play a less critical bridging role. In contrast, social media and print media exhibit lower centrality scores, indicating their peripheral influence. The findings suggest that while traditional information sources remain dominant, digital platforms are emerging as valuable tools for information dissemination highlighting the need for strategies that strengthen traditional channels while integrating digital tools to enhance the efficiency and reach of agricultural information networks. This research contributes to a deeper understanding of the complex information ecosystem in agriculture, offering insights for policymakers and stakeholders aiming to optimize information flows within the sector.

Predicting stroke volume variation using central venous pressure waveform: a deep learning approach

Objective. This study evaluated the predictive performance of a deep learning approach to predict stroke volume variation (SVV) from central venous pressure (CVP) waveforms.Approach. Long short-term memory (LSTM) and the feed-forward neural network were sequenced to predict SVV using CVP waveforms obtained from the VitalDB database, an open-source registry. The input for the LSTM consisted of 10 s CVP waveforms sampled at 2 s intervals throughout the anesthesia duration. Inputs of the feed-forward network were the outputs of LSTM and demographic data such as age, sex, weight, and height. The final output of the feed-forward network was the SVV. The performance of SVV predicted by the deep learning model was compared to SVV estimated derived from arterial pulse waveform analysis using a commercialized model, EV1000.Main results. The model hyperparameters consisted of 12 memory cells in the LSTM layer and 32 nodes in the hidden layer of the feed-forward network. A total of 224 cases comprising 1717 978 CVP waveforms and EV1000/SVV data were used to construct and test the deep learning models. The concordance correlation coefficient between estimated SVV from the deep learning model were 0.993 (95% confidence interval, 0.992-0.993) for SVV measured by EV1000.Significance. Using a deep learning approach, CVP waveforms can accurately approximate SVV values close to those estimated using commercial arterial pulse waveform analysis.

Reimagining otitis media diagnosis: A fusion of nested U-Net segmentation with graph theory-inspired feature set

Otitis media (OM) is a common infection or inflammation of the middle ear causing conductive hearing loss that primarily affects children and may delay speech, language, and cognitive development. OM can manifest itself in different forms, and can be diagnosed using (video) otoscopy (visualizing the tympanic membrane) or (video) pneumatic otoscopy and tympanometry. Accurate diagnosis of OM is challenging due to subtle differences in otoscopic features. This research aims to develop an automated computer-aided design (CAD) system to assist clinicians in diagnosing OM using otoscopy images. The ground truths, generated manually and validated by otolaryngologists, are utilized to train the proposed nested U-Net++ model. Ten clinically relevant gray level co-occurrence matrix (GLCM) and morphological features were extracted from the segmented Region of Interest (ROI) and validated for OM classification based on a statistical significance test. These features serve as input for a Graph Neural Network (GNN) model, the base model in our research. An optimized GNN model is proposed after ablation study of the base model. Three datasets, one private dataset, and two public ones have been used, where the private dataset is utilized for both training and testing, and the public datasets are used to test the robustness of the proposed GNN model only. The proposed GNN model obtained the highest accuracy in diagnosing OM: 99.38 %, 93.51 %, and 91.38 % for the private dataset, public dataset1, and public dataset2, respectively. The proposed methodology and results of this research might enhance clinicians' effectiveness in diagnosing OM.

Backpropagation-Based Learning Techniques for Deep Spiking Neural Networks: A Survey.

With the adoption of smart systems, artificial neural networks (ANNs) have become ubiquitous. Conventional ANN implementations have high energy consumption, limiting their use in embedded and mobile applications. Spiking neural networks (SNNs) mimic the dynamics of biological neural networks by distributing information over time through binary spikes. Neuromorphic hardware has emerged to leverage the characteristics of SNNs, such as asynchronous processing and high activation sparsity. Therefore, SNNs have recently gained interest in the machine learning community as a brain-inspired alternative to ANNs for low-power applications. However, the discrete representation of the information makes the training of SNNs by backpropagation-based techniques challenging. In this survey, we review training strategies for deep SNNs targeting deep learning applications such as image processing. We start with methods based on the conversion from an ANN to an SNN and compare these with backpropagation-based techniques. We propose a new taxonomy of spiking backpropagation algorithms into three categories, namely, spatial, spatiotemporal, and single-spike approaches. In addition, we analyze different strategies to improve accuracy, latency, and sparsity, such as regularization methods, training hybridization, and tuning of the parameters specific to the SNN neuron model. We highlight the impact of input encoding, network architecture, and training strategy on the accuracy-latency tradeoff. Finally, in light of the remaining challenges for accurate and efficient SNN solutions, we emphasize the importance of joint hardware-software codevelopment.

Spatial–temporal prediction model for unsteady near-wall flow around cylinder based on hybrid neural network

A hybrid neural network based on Densely Connected Convolutional Networks (DenseNet), Convolutional Long Short-Term Memory Neural Network (ConvLSTM), and Deconvolutional Neural Network (DeCNN) is employed to predict unsteady flow fields. The utilization of DenseNet makes the model more compact and makes the prediction of three-dimensional flow affordable. The ConvLSTM is implemented to predict multiple future time steps which improves prediction efficiency. The proposed model transforms the time sequences of velocity and pressure fields into uniform spatial–temporal topology as input and captures nonlinear feature information in the spatial–temporal domain. Numerical simulations are conducted for the flow around cylinder at different Reynolds numbers and the near-wall flow around cylinder with different gap ratios, and training samples for the neural network inputs are established. The predicted results are compared with the numerical simulation results, showing good agreement. From the prediction cycle, it can be seen that good prediction results can be maintained in the first three prediction cycles. The prediction results of the three-dimensional unsteady flow around a cylinder near a plane wall, exhibit remarkable accuracy, successfully capturing the evolution of turbulent vortex structures. This signifies that the prediction model is highly effective in capturing the spatial–temporal variations of complex unsteady flows.