Cascade Neural Network Research Articles

A Fast Prediction Method for the Electromagnetic Response of the LTE-R System Based on a PSO-BP Cascade Neural Network Model

In this paper, a fast prediction model of the electromagnetic response of the LTE-R (Long Term Evolution for Railway) communication system based on a cascade neural network is developed to quickly analyze the impact of pantograph arcing on the LTE-R system during vehicle operation. In order to obtain the coupling disturbance level of the LTE-R antenna port, this model uses the cascade neural network based on the PSO-BP (Particle Swarm Optimization of Back Propagation Neural Network) algorithm to quickly solve the coupling coefficient of the pantograph arcing measurement probe and the antenna port. A two-stage BP neural network model is used to train both the simulation data and measurement data, and the results are validated by field tests. Finally, the antenna port coupling interference is added to the LTE-R system to analyze the impact of pantograph arcing on the quality of communication transmission. Analysis and tests demonstrate that pantograph arcing can lead to an increase in the BLER of LTE-R systems at 400 MHz, affecting system performance and transmission efficiency.

Thermal power plants pollution assessment based on deep neural networks, remote sensing, and GIS: A real case study in Iran

To investigate the impact of the Bandar Abbas thermal power plant on the waters of the Persian Gulf coast, a combination of satellite images and ground data was utilized to determine the Sea Surface Temperature (SST) as a thermal index, Total Organic Carbon (TOC) and Chemical Oxygen Demand (COD) as biological indices. Additionally, measurements of SO2, O3, NO2, CO2, CO, and CH4 values in the atmosphere were taken to determine the plant's impact on air pollution. Temperature values of the water for different months were predicted using Long Short-Term Memory (LSTM), Support Vector Regression (SVR), and Cascade neural networks. The results indicate that the waters near thermal power plants exhibit the highest temperatures in July and September, with temperatures reaching approximately 50 °C. Furthermore, the SST values were found to be strongly correlated with ecological indices. The Multiple Linear Regression (MLR) analysis revealed a strong correlation between the temperature and TOC, COD, and O2 in water (RTOC2=0.98), RO22=−0.89, RCOD2=0.87 and O3, NO3, CO2, and CO in the air (RO32=0.99,RNO32=0.97,RCO22=0.95,RCO2=0.96). Finally, the results demonstrate that the LSTM method exhibited high accuracy in predicting the water temperature (R2 = 0.98).

Simulating and Comparing CO2/CH4 Separation Performance of Membrane-Zeolite Contactors by Cascade Neural Networks.

Separating carbon dioxide (CO2) from gaseous streams released into the atmosphere is becoming critical due to its greenhouse effect. Membrane technology is one of the promising technologies for CO2 capture. SAPO-34 filler was incorporated in polymeric media to synthesize mixed matrix membrane (MMM) and enhance the CO2 separation performance of this process. Despite relatively extensive experimental studies, there are limited studies that cover the modeling aspects of CO2 capture by MMMs. This research applies a special type of machine learning modeling scenario, namely, cascade neural networks (CNN), to simulate as well as compare the CO2/CH4 selectivity of a wide range of MMMs containing SAPO-34 zeolite. A combination of trial-and-error analysis and statistical accuracy monitoring has been applied to fine-tune the CNN topology. It was found that the CNN with a 4-11-1 topology has the highest accuracy for the modeling of the considered task. The designed CNN model is able to precisely predict the CO2/CH4 selectivity of seven different MMMs in a broad range of filler concentrations, pressures, and temperatures. The model predicts 118 actual measurements of CO2/CH4 selectivity with an outstanding accuracy (i.e., AARD = 2.92%, MSE = 1.55, R = 0.9964).

A cascade neural network methodology for fault detection and diagnosis in solar thermal plants

Detecting and isolating faults in collector fields of solar thermal power plants is a crucial and challenging task. The system variables in the collector area are highly coupled, which can lead to a high misclassification rate. For this reason, it becomes necessary to combine knowledge of systems engineering with machine learning techniques that unravel the complex dynamics that govern the systems using historical data. Furthermore, the performance of a solar thermal plant is highly dependent on solar irradiance which changes during the day and is subject to perturbations caused by clouds and other atmospheric conditions. Detecting the fault requires using techniques that cope with the disturbances in solar irradiance.In this work, real irradiance profiles with many types of clouds are used. First, a model-based fault detector is applied, obtaining an accuracy of over 89% for all test irradiances. Then, different machine learning techniques are compared: static neural networks with and without decoupling strategy, dynamic neural networks, dynamic neural networks in cascade, classification trees, random forests, radial basis function networks, and self-organizing maps. The combination of neural networks was the only method that obtained a total accuracy of over 73% and F1-scores over 50% for all the test irradiance profiles.

Multivariable identification of membrane fouling based on compacted cascade neural network

The system consists of three modules: data acquisition, feature extraction, and online prediction. The data acquisition module obtains the value of process variables from the sensor and transfers the acquired variables to the database via PLC. The feature extraction module is designed to filter the collected process data. In this module, multiple related variables are selected as the preselected variables. Then the PLS method is used to reduce the dimension of the preselected variables to obtain the feature variables with a high correlation with the predicted variables. Finally, the online prediction module based on a cascade neural network predicts membrane fouling, membrane integrity, and membrane life for clean decisions. • A multivariable identification model is construted. • Cascade neural network is proposed to predict multiple variables to avoid the interference of overlap inputs. • Unsupervised pretraining algorithm is used to initialize the structure of cascade neural network. • Because the parameters are affected by multiple errors, the hierarchical learning algorithm is used to improve the identification accuracy. The membrane fouling phenomenon, reflected with various fouling characterization in the membrane bioreactor (MBR) process, is so complicated to distinguish. This paper proposes a multivariable identification model (MIM) based on a compacted cascade neural network to identify membrane fouling accurately. Firstly, a multivariable model is proposed to calculate multiple indicators of membrane fouling using a cascade neural network, which could avoid the interference of the overlap inputs. Secondly, an unsupervised pretraining algorithm was developed with periodic information of membrane fouling to obtain the compact structure of MIM. Thirdly, a hierarchical learning algorithm was proposed to update the parameters of MIM for improving the identification accuracy online. Finally, the proposed model was tested in real plants to evaluate its efficiency and effectiveness. Experimental results have verified the benefits of the proposed method.

Garbage classification based on a cascade neural network

Most existing methods of garbage classification utilize transfer learning to acquire acceptable performance. They focus on some smaller categories. For example, the number of the dataset is small or the number of categories is few. However, they are hardly implemented on small devices, such as a smart phone or a Raspberry Pi, because of the huge number of parameters. Moreover, those approaches have insufficient generalization capability. Based on the aforementioned reasons, a promising cascade approach is proposed. It has better performance than transfer learning in classifying garbage in a large scale. In addition, it requires less parameters and training time. So it is more suitable to a potential application, such as deployment on a small device. Several commonly used backbones of convolutional neural networks are investigated in this study. Two different tasks, that is, the target domain being the same as the source domain and the former being different from the latter, are conducted besides. Results indicate with ResNet101 as the backbone, our algorithm outperforms other existing approaches. The innovation is that, as far as we know, this study is the first work combining a pre-trained convolutional neural network as a feature extractor with extreme learning machine to classify garbage. Furthermore, the training time and the number of trainable parameters is significantly shorter and less, respectively.

Software Effort Estimation Using Cascade Neural Network Optimized Based on Modified Particle Swarm Optimization (MPSO-CNN)

Software Effort Estimation Using Cascade Neural Network Optimized Based on Modified Particle Swarm Optimization (MPSO-CNN)

Comparison of Various Prediction Model for Biodiesel Cetane Number using Cascade-Forward Neural Network

Cetane number (CN) is one of the important fuel properties of diesel fuels. It is a measurement of the ignition quality of diesel fuel. Numerous studies have been published to predict the CN of biodiesels. More recently, the utilization of soft computing methods such as artificial neural networks (ANN) has received considerable attention as a prediction tool. However, most studies in the use of ANN for estimating the CN of biodiesels have only used one algorithm to train a small number of datasets. This study aims to predict the CN of 63 biodiesels based on the fatty acid methyl esters (FAME) composition by developing an ANN model that was trained with 10 different algorithms. To the best of our knowledge, this is the first study to predict the CN of biodiesels using numerous ANN training algorithms utilizing sizeable datasets. Results revealed that the ANN model trained with Levenberg-Marquardt gave the highest prediction accuracy. LM algorithm successfully predicted the CN of biodiesels with the highest correlation and determination coefficient (R = 0.9615, R2 = 0.9245) as well as the lowest errors (MAD = 2.0804, RMSE = 3.1541, and MAPE = 4.2971). Hence, the Cascade neural network trained with the LM algorithm could be considered a promising alternative to the empirical correlations for predicting biodiesel’s CN.

Hierarchical attention neural network for information cascade prediction

Online social networking platforms have drastically facilitated the phenomenon of information cascades, making cascade prediction an important task for both researchers and practitioners. In cascade propagation paths, influential users can often attract more attention. Moreover, the community structure formed by users with similar interests creates information redundancy, thereby restricting the propagation process. These two features greatly affect the growth of the cascades. This paper investigates the incremental prediction of cascades during a future time period based on the evolution of cascades in early stages and proposes a neural network framework with hierarchical attention mechanisms, named Hierarchical Attention Cascade Neural Network (CasHAN). This network has a node-level attention mechanism based on user influence and a sequence-level attention mechanism based on community redundancy. User influence considers the user’s own attributes and the influence feedback provided by neighbors, while community redundancy measures information redundancy characteristics that limit cascade propagation. Through hierarchical attention mechanisms, the effective combination of these two features improves the accuracy of cascade increment prediction. Extensive experiments on real-world datasets demonstrate that our approach outperforms other state-of-the-art prediction methods.

Development of neural network controller for control of debutanization process in rectification column of gas-fractionating plant

In process systems, stabilizing parameters and bringing the process to technological mode is a rather difficult task. Control and control of parameters in manual mode does not produce effective results, and for this reason the introduction of an improved control system is required to eliminate human factor errors. Currently, the best efficiency is shown by neural networks, which are used for an advanced control and prediction system. The purpose of the study: in this project, a neural network controller is being developed to control the debutanization process of the gas fractionation plant in the Matlab Simulink environment. Cascade neural networks consisting of a certain number of layers and neurons are used to develop the project. During the work, a cascade neural network was selected, in which the parameters of the technological system were trained.

MSCN-NET: Multi-stage cascade neural network based on attention mechanism for Čerenkov luminescence tomography

Čerenkov luminescence tomography (CLT) is a highly sensitive and promising technique for three-dimensional non-invasive detection of radiopharmaceuticals in living organisms. However, the severe photon scattering effect causes ill-posedness of the inverse problem, and the results of CLT reconstruction are still unsatisfactory. In this work, a multi-stage cascade neural network is proposed to improve the performance of CLT reconstruction, which is based on the attention mechanism and introduces a special constraint. The network cascades an inverse sub-network (ISN) and a forward sub-network (FSN), where the ISN extrapolates the distribution of internal Čerenkov sources from the surface photon intensity, and the FSN is used to derive the surface photon intensity from the reconstructed Čerenkov source, similar to the transmission process of photons in living organisms. In addition, the FSN further optimizes the reconstruction results of the ISN. To evaluate the performance of our proposed method, numerical simulation experiments and in vivo experiments were carried out. The results show that compared with the existing methods, this method can achieve superior performance in terms of location accuracy and shape recovery capability.

Investigation of Nonlinear Vibrational Analysis of Circular Sector Oscillator by Using Cascade Learning

This paper analyzed the model of swinging oscillation of a solid circular sector arising in hydrodynamical machines, electrical engineering, heat transfer applications, and civil engineering. Nonlinear differential equations govern the mathematical model for frequency oscillation of the system. Furthermore, a computational strength of Cascade neural networks (CNNs) is utilized with backpropagated Levenberg–Marquardt (BLM) algorithm to study the oscillations in angular displacement θ , velocity θ ′ , and acceleration θ ″ . A data set for the supervised learning of the CNN-BLM algorithm for different angles α and radius R are generated by Runge–Kutta (RK-4) method. The BLM algorithm further interprets the dataset with log-sigmoid as an activation function for the solutions’ validation, testing, and training. The results obtained by the design scheme are compared with Akbari–Ganji’s (AG) method. The rapid convergence and quality of the solutions are validated through performance indicators such as mean absolute deviations (MAD), root means square error, and error in Nash–Sutcliffe efficiency (ENSE). The statistics demonstrate the design scheme’s applicability and efficiency to highly singular nonlinear problems.

Cascade neural approximating for few-shot super-resolution photoacoustic angiography

High-resolution photoacoustic angiography images are reconstructed from undersampled images with the help of a super-resolution deep neural network, enhancing the ability of the photoacoustic angiography systems to image dynamic processes in living tissues. However, image degradations are difficult to estimate due to a lack of knowledge of the point spread function and noise sources, resulting in poor generalization capability of the trained super-resolution model. In this work, a high-order residual cascade neural network was developed to reconstruct high-resolution vascular images, which is a neural approximating approach used to remove image degradations of photoacoustic angiography. To handle overfitting in training super-resolution model with a limited dataset, we proposed a BicycleGAN based image synthesis method in data preparation, achieving a strong regularization by forging realistic photoacoustic vascular images that act to essentially increase the training dataset. The quantitative analysis of the reconstructed results shows that the high-order residual cascade neural network surpassed the other residual super-resolution neural networks. Most importantly, we demonstrated that the generalized model could be achieved despite the limited training dataset, promising to be a methodology for few-shot super-resolution photoacoustic angiography.

Heat transfer analysis of an inclined longitudinal porous fin of trapezoidal, rectangular and dovetail profiles using cascade neural networks

Heat transfer analysis of an inclined longitudinal porous fin of trapezoidal, rectangular and dovetail profiles using cascade neural networks

A secure finger vein verification and authentication scheme for banking network

SummaryThe development of electronic exchanges has created a more important interest in fast and accurate customer IDs and verification. Structures, bank records, and PC framework often use personally identifiable access code IDs and test numbers (PINs) for reliable status. Biometric verification requires a level of security, unchanged structure, spoofing, and reliability. In this research work, a biometric authentication system was developed to provide security for access to the ATM network. Honest technology used for security purposes to provide efficient ATM network security is a finger vein. This method undergoes image processing technology such as preprocessing, feature extraction, and classification to locate the finger vein region for accessing the ATM network. Finally, the cascade neural network acts as a classifier method to differentiate registered and unregistered user. The performance analysis of this method is simulated by MATLAB and the process is evaluated to measure the efficient security of the network. The dataset is segregated into two stages; one for training and other for testing. The features extracted with the help of hidden layers with neurons. The matched finger vein is compared to check the accuracy.

Intelligent deep learning based product image recommendation and classification system

Recently, recommendation system (RS) has gained significant attention in several industries and business sectors. At the same time, image recommendation is also found helpful to determine the relevant objects that exist in the form of image. It is mainly based on the extraction of different features and utilize it to get recommended outcomes. The recently developed deep learning (DL) models can be applied to design effective product image recommendation and classification systems. In this aspect, this paper designs an intelligent deep learning enabled product image recommendation and classification (IDL-PIRC) system. The proposed IDL-PIRC technique aims to examine the input image to recommend and classify products based on the user query. In addition, the proposed IDL-PIRC technique involves Gaussian filtering (GF) based pre-processing to eradicate the existence of noise exist in it. Moreover, the fusion-based feature extraction technique uses the Grey-Level Run Length Matrix (GLRLM) and Residual Network (ResNet152) models. Furthermore, the kNN based ranking approach is employed for product recommendation and cascaded neural network (CNN) is utilized for product classification. A wide range of simulations take place and the results are inspected interms of different evaluation parameters.

AUTO-ASSISTANCE SYSTEM FOR VISUALLY IMPAIRED PEOPLE USING DEEP LEARNING

World Health Organization (WHO) outlined that there are two eighty-five million visually disabled worldwide. Among them thirty-nine million people are completely blind. One of the difficult activities that could be conducted by visually impaired is object detection which could be implemented using Machine Learning. It is an approach towards Artificial Intelligence that provides system the capacity for natural learning and development from experience without specifically programmed. It provides computer vision to the system which make decisions based on training algorithms. The chief goal of this research paper is to develop an object detection system to assist totally blind individual to manage their activities independently. Paper also compares different object detection algorithms like Haar Cascade and Convolutional Neural Network with yolov5(CNN). Haar Cascade classifier is a basic face detection algorithm which could also be trained to detect different objects whereas convolutional neural network falls under deep learning approach which could be employed for object recognition. The custom dataset is created with 2300 images consisting of 3 different classes. This comparison is being executed to find the yolov5 as a suitable algorithm for this system from the aspect of accuracy for real time scenario. Keywords- Object Detection, Computer Vision, Deep learning, Feature Extraction and Recognition , Convolution neural network.

Transfer learning for cognitive reserve quantification

Cognitive reserve (CR) has been introduced to explain individual differences in susceptibility to cognitive or functional impairment in the presence of age or pathology. We developed a deep learning model to quantify the CR as residual variance in memory performance using the Structural Magnetic Resonance Imaging (sMRI) data from a lifespan healthy cohort. The generalizability of the sMRI-based deep learning model was tested in two independent healthy and Alzheimer's cohorts using transfer learning framework.Structural MRIs were collected from three cohorts: 495 healthy adults (age: 20-80) from RANN, 620 healthy adults (age: 36-100) from lifespan Human Connectome Project Aging (HCPA), and 941 adults (age: 55-92) from Alzheimer's Disease Neuroimaging Initiative (ADNI). Region of interest (ROI)-specific cortical thickness and volume measures were extracted using the Desikan-Killiany Atlas. CR was quantified by residuals which subtract the predicted memory from the true memory. Cascade neural network (CNN) models were used to train RANN dataset for memory prediction. Transfer learning was applied to transfer the T1 imaging-based model from source domain (RANN) to the target domains (HCPA or ADNI).The CNN model trained on the RANN dataset exhibited strong linear correlation between true and predicted memory based on the T1 cortical thickness and volume predictors. In addition, the model generated from healthy lifespan data (RANN) was able to generalize to an independent healthy lifespan data (HCPA) and older demented participants (ADNI) across different scanner types. The estimated CR was correlated with CR proxies such education and IQ across all three datasets.The current findings suggest that the transfer learning approach is an effective way to generalize the residual-based CR estimation. It is applicable to various diseases and may flexibly incorporate different imaging modalities such as fMRI and PET, making it a promising tool for scientific and clinical purposes.

Computational Intelligence-Based Melanoma Detection and Classification Using Dermoscopic Images.

Melanoma is a kind of skin cancer caused by the irregular development of pigment-producing cells. Since melanoma detection efficiency is limited to different factors such as poor contrast among lesions and nearby skin regions, and visual resemblance among melanoma and non-melanoma lesions, intelligent computer-aided diagnosis (CAD) models are essential. Recently, computational intelligence (CI) and deep learning (DL) techniques are utilized for effective decision-making in the biomedical field. In addition, the fast-growing advancements in computer-aided surgeries and recent progress in molecular, cellular, and tissue engineering research have made CI an inevitable part of biomedical applications. In this view, the research work here develops a novel computational intelligence-based melanoma detection and classification technique using dermoscopic images (CIMDC-DIs). The proposed CIMDC-DI model encompasses different subprocesses. Primarily, bilateral filtering with fuzzy k-means (FKM) clustering-based image segmentation is applied as a preprocessing step. Besides, NasNet-based feature extractor with stochastic gradient descent is applied for feature extraction. Finally, the manta ray foraging optimization (MRFO) algorithm with a cascaded neural network (CNN) is exploited for the classification process. To ensure the potential efficiency of the CIMDC-DI technique, we conducted a wide-ranging simulation analysis, and the results reported its effectiveness over the existing recent algorithms with the maximum accuracy of 97.50%.

Evolving Hybrid Generalized Space-Time Autoregressive Forecasting with Cascade Neural Network Particle Swarm Optimization

Background: The generalized space-time autoregressive (GSTAR) model is one of the most widely used models for modeling and forecasting time series and location data. Methods: In the GSTAR model, there is an assumption that the research locations are heterogeneous. In addition, the differences between these locations are shown in the form of a weighting matrix. The novelty of this paper is that we propose the hybrid time-series model of GSTAR uses the cascade neural network and obtains the best parameters from particle swarm optimization. Results and conclusion: This hybrid model provides a high accuracy value for forecasting PM2.5, PM10, NOx, and SO2 with high accuracy forecasting, which is justified by a mean absolute percentage error (MAPE) accuracy of around 0.01%.