Deep Neural Network Algorithm Research Articles

Implementation of specifically designed deep neural networks for the prediction and optimization of tensile properties of aluminum-copper alloy

Alloys are engineered materials aimed at enhancing mechanical properties. Extensive research has focused on identifying the optimal metal composition for alloys with superior tensile strength. This study validates the stiffness and strength values of an aluminum-copper alloy through a comparison with a molecular dynamics simulation. Subsequently, 100 data points were obtained from the simulation, and a deep neural network (DNN) with three hidden layers was employed. The DNN was trained, tested, and its structure optimized using the Taguchi design of experiment. The proposed DNN structures successfully predicted the maximum values of the stiffness and strength, which were further verified using molecular dynamics simulation. Notably, the results demonstrated the complete reliability of the Taguchi-designed DNN algorithm in this application.

Optuna-DFNN: An Optuna framework driven deep fuzzy neural network for predicting sintering performance in big data

Sintering performance index can reflect the quality of sintered ore, and the level of sintered ore performance directly affects the stability of blast furnace production. Accurate prediction of sinter performance metrics is essential to optimise sinter quality, improve production efficiency and reduce energy consumption. Based on the sintering big data, this paper proposes the Optuna-DFNN model by combining the fuzzy neural network algorithm, the deep neural network algorithm and the Optuna framework to address the problems of difficulty in tuning the parameter, lack of self-learning ability, and poor generalisation of the model of the traditional method when faced with the input of multiple parameters. The paper predicts the four prediction indexes of yield rate, return fines ratio, drum index and RDI+3.15 respectively. Among them, the relative error of the prediction of the yield rate is within 1.3 %, the relative error of the prediction of the return fines ratio is within 2.8 %, the relative error of the prediction of the drum index is within 0.32 %, and the relative error of the prediction of the RDI+3.15 is within 1.4 %. The results indicate that the Optuna-DFNN model has better performance in predicting sintering performance metrics. Based on the Optuna-DFNN sintered ore performance prediction model, the quality of sintered ore can be accurately predicted, the sintering process parameters can be adjusted in a timely manner, the sintering process can be optimized, and the utilization coefficient of sintering can be improved. Provide a reference for the optimisation of production processes, with a positive impact on cost reduction, environmental protection and increased sustainability of production.

A deep neural network for predicting soil texture using airborne radiometric data

The ternary nature of soil texture, defined by its proportions of clay, silt, and sand, makes it challenging to predict through linear regression models from other soil attributes and auxiliary variables. The most promising results in this field have been recently achieved by Machine Learning methods which are able to derive non-linear, non-site-specific models to predict soil texture. In this paper we present a method of constructing a pair of Deep Neural Network (DNN) algorithms that can predict clay and sand soil contents from Airborne Gamma Ray Spectrometry data of K and Th ground abundances.We tested the algorithm's hyperparameters through various configurations to optimize the DNNs' performance, effectively avoiding underfitting and overfitting of the models. This led to the creation of a high-resolution 20 m × 20 m soil texture map from dense AGRS data, significantly refining the previous map's granularity. The application of the obtained DNN models to unseen sites can be supported by future training on additional textural classes.

IOT based Waste Segregation Using Deep Learning

In the rapidly urbanizing world, effective waste management is crucial to maintain environmental sustainability. This project presents an innovative solution utilizing Internet of Things (IoT) technology and deep learning Convolutional Neural Networks (CNN) algorithm for automated waste segregation. We propose a smart waste segregation system equipped with sensors to detect the level of waste in bins. The system utilizes a CNN model trained using Deep Learning to classify waste items into different categories, ensuring proper segregation. When a bin reaches its full capacity, the IoT sensors trigger an alert mechanism, sending real-time notifications to the concerned authorities or waste management personnel. This proactive approach not only optimizes waste collection but also promotes recycling by ensuring the segregated waste reaches the appropriate recycling facilities. The integration of IoT devices and deep learning algorithms enhances the efficiency of waste management processes, reducing environmental pollution and promoting a sustainable future. Key Words: IoT, waste management, deep learning, CNN algorithm, waste segregation, environmental sustainability, real-time monitoring, sensor technology, recycling, smart cities, automation.

Boosting Clear Cell Renal Carcinoma-Specific Drug Discovery Using a Deep Learning Algorithm and Single-Cell Analysis.

Clear cell renal carcinoma (ccRCC), the most common subtype of renal cell carcinoma, has the high heterogeneity of a highly complex tumor microenvironment. Existing clinical intervention strategies, such as target therapy and immunotherapy, have failed to achieve good therapeutic effects. In this article, single-cell transcriptome sequencing (scRNA-seq) data from six patients downloaded from the GEO database were adopted to describe the tumor microenvironment (TME) of ccRCC, including its T cells, tumor-associated macrophages (TAMs), endothelial cells (ECs), and cancer-associated fibroblasts (CAFs). Based on the differential typing of the TME, we identified tumor cell-specific regulatory programs that are mediated by three key transcription factors (TFs), whilst the TF EPAS1/HIF-2α was identified via drug virtual screening through our analysis of ccRCC's protein structure. Then, a combined deep graph neural network and machine learning algorithm were used to select anti-ccRCC compounds from bioactive compound libraries, including the FDA-approved drug library, natural product library, and human endogenous metabolite compound library. Finally, five compounds were obtained, including two FDA-approved drugs (flufenamic acid and fludarabine), one endogenous metabolite, one immunology/inflammation-related compound, and one inhibitor of DNA methyltransferase (N4-methylcytidine, a cytosine nucleoside analogue that, like zebularine, has the mechanism of inhibiting DNA methyltransferase). Based on the tumor microenvironment characteristics of ccRCC, five ccRCC-specific compounds were identified, which would give direction of the clinical treatment for ccRCC patients.

NOVEL METHOD VITAMIN DEFICIENCY DETECTION USING ALEXNET DNN ALGORITHM

Vitamin deficiencies can have significant impacts on overall health and well-being. Early detection plays a crucial role in preventing complications and improving outcomes. However, traditional methods for detecting deficiencies can be time-consuming and costly. This project aims to develop a novel method for detecting vitamin deficiencies using the AlexNet DNN algorithm, a powerful deep learning model for image classification. The purpose of this project is to explore the feasibility of using image analysis and deep learning techniques to detect vitamin deficiencies accurately and efficiently. The objectives include improving the accuracy of detection, reducing false positives and negatives, and developing a reliable and accessible tool for early detection. To achieve our objectives, we will gather a large dataset of images depicting various vitamin deficiencies. These images will be preprocessed to enhance features and reduce noise. The AlexNet DNN algorithm will be trained on this dataset, learning to recognize patterns and features associated with different deficiencies. The algorithm will undergo rigorous testing and evaluation to ensure its effectiveness. KEYWORDS— Vitamins, Deficiency, AlexNet, Deep Neural Network(DNN), Effectiveness

Hate Speech Detection

Abstract: The rise in popularity of microblogging sites such as Facebook, Instagram, and Twitter has resulted in more people from different backgrounds indirectly communicating with one another. Our study aims to design an autonomous Deep Neural Network (DNN) algorithm for social media hate speech detection to tackle this problem. Using cutting-edge Natural Language Processing (NLP) techniques, the objective is to build a strong system that can recognize and categorize hate speech material in text data with accuracy. Our DNN algorithm allows for the real-time detection and moderation of offensive information, providing a proactive strategy against online hate speech. With the deployment of this technology, everyone will be able to access a safer and more welcoming online environment.

Automatic recognition of Rice Plant leaf diseases detection using deep neural network with improved threshold neural network

The farming industry widely requires automatic detection and analysis of rice diseases to avoid wasting financial and other resources, reduce yield loss, improve processing efficiency, and obtain healthy crop yields. The progress made in deep learning techniques has significantly impacted agricultural disease diagnostics. Rice Plant leaf diseases can have severe adverse effects on crop yield, and proper diagnosis of Rice Plant diseases is vital to avoid these effects. Most of the existing methods do not adequately detect the leaf images and cannot find the condition accurately. Previously, there were some problems with leaf segmentation. Complex disease stages are efficient but computationally time-consuming, and segmentation needs to be more accurate, affordable, and reliable. To overcome the issues, the proposed method of Deep Spectral Generative Adversarial Neural Network (DSGAN2) with Improved Artificial Plant Optimization for Rice Plant leaf disease detection. Initially, we fed into the input of healthy and non-healthy leaves from the collected dataset. Then, we apply an Improved Threshold Neural Network (ITNN) method to enhance the image quality. Next, it uses a Segmentation using a Segment Multiscale Neural Slicing (SMNS) algorithm to identify the support-intensive color saturation based on the enhanced image. After that, the Spectral Scaled Absolute Feature Selection (S2AFS) method is applied to select optimal features and the Closest Weight from segmented Rice Plant leaves. Social Spider Optimization to select the Feature with the Closest Weight (S2O-FCW) algorithm for analysis of the feature weight values. Finally, the proposed Soft-Max Logistic Activation Function with Deep Spectral Generative Adversarial Neural Network (DSGAN2) algorithm detects Rice Plant disease based on selected features. The proposed system Deep Spectral Generative Adversarial Neural Network (DSGAN2) produces a decreasing false rate compared to the existing system of ACPSOSVM-Dual Channels Convolutional Neural Network (APS-DCCNN) is 55.2 %, Alex Net is 50.4 %, and Convolutional Neural Network (CNN) is 49.5 %.

The development of a model for the threat detection system with the use of machine learning and neural network methods

This study examines the development of a model for the threat detection system with the use of machine learning and neural network methods. The fast development of Internet technologies has led to the appearance of many digital systems and platforms. However, despite the impressive technological progress, another side also emerged in the spread of a massive number of different cyber threats. Although various ways have been created to detect and prevent them, the threats are also developing and becoming more complex each year. Therefore, new system defense and data protection methods using machine and deep learning approaches have been proposed recently. The methods based on these approaches have proved to be especially effective in the wave of new Artificial Intelligence applications. In this paper, a threat detection system has been designed to disclose different kinds of threats while maintaining the security, confidentiality, and availability of the computer system. The development of machine learning models for detecting DDoS and man-in-the-middle attacks, Structured Query Language (SQL) injections, phishing, and malware was examined. The data scaling, feature selection, feature extraction, and classification steps were also thoroughly described. Naïve Bayes, Logistic Regression, Decision Tree, Random Forest, XGBoost, CatBoost, and Deep Neural Network algorithms were utilized for training the cyber threat detection models. The experimental results evaluated all the models using accuracy, precision, recall, and F1-score metrics. The best models achieved scores in the range of 0.90 to 1.00.

Prevention of Thermal Crack in Steel Slab Using Neural Networks Model to Predict Impact Absorption Energy

To prevent thermal cracks on steel slabs during the stacking, scarfing, and grinding processes after continuous casting, an optimized machine learning (ML) model to predict the impact absorption energy (EIA) of steel slabs is developed. A total of 1,421 experimental EIA data are collected from Charpy impact tests at slab temperatures 25 ≤ TS ≤ 400 °C, then ML models that considered 15 steel‐component variables and temperatures are developed. Four ML models are developed and their accuracies are compared. The optimized deep neural network algorithm predicts EIA most accurately with root mean squared error of 10.82 J and coefficient of determination (R2) of 0.992. Then the predicted EIA is used as the criterion to predict the occurrence of thermal cracks in slabs that are actually produced by a continuous casting process. At TS = 250 °C, cracking does not occur when the steel has predicted EIA > 175 J. The use of the developed model to predict EIA can prevent the formation of thermal cracks in slabs produced by continuous casting, and enable optimization of the cooling method and scarfing methods that precede the next process after continuous casting of slabs.

Algorithmic Foundation of Federated Learning with Sequential Data

The current analysis of federated optimization algorithms for training deep neural networks assumes that the data is non-sequential (e.g., images), which incurs a smooth loss objective. In contrast, edge devices generate lots of sequential data every day, where these sequences exhibit significant sequential correlation at different time stamps (e.g., text messages). In order to learn from such sequential data, people typically use a class of neural networks that is inherently nonsmooth, with a potentially unbounded smoothness parameter. Examples include recurrent neural networks, long-short-term memory networks, and transformers. It remains unclear how to design provably efficient algorithms for training these neural networks to learn from sequential data. My goal is to lay the algorithmic foundation of federated learning with sequential data, which contributes novel algorithms for learning from a range of real-world sequential data (e.g., natural language, electronic health record, transportation, time series, etc.) using state-of-the-art deep neural networks. In this talk, I will first motivate the problem by showing that the transformer, which is widely used for sequential data learning, has an unbounded smooth landscape. Then, I will introduce provably efficient federated deep learning algorithms in the presence of unbounded smoothness. In particular, I will introduce a few efficient algorithms for various settings of federated learning, including homogeneous data, heterogeneous data, and partial client participation. The main result is twofold. First, we show that the designed algorithms provably small computational and communication complexities. Second, we establish fundamental hardness results in the unbounded smoothness setting. Ultimately, I will discuss the future challenges of extending our research framework from small-scale neural networks to large language models.

Learning Korobov Functions by Correntropy and Convolutional Neural Networks.

Combining information-theoretic learning with deep learning has gained significant attention in recent years, as it offers a promising approach to tackle the challenges posed by big data. However, the theoretical understanding of convolutional structures, which are vital to many structured deep learning models, remains incomplete. To partially bridge this gap, this letter aims to develop generalization analysis for deep convolutional neural network (CNN) algorithms using learning theory. Specifically, we focus on investigating robust regression using correntropy-induced loss functions derived from information-theoretic learning. Our analysis demonstrates an explicit convergence rate for deep CNN-based robust regression algorithms when the target function resides in the Korobov space. This study sheds light on the theoretical underpinnings of CNNs and provides a framework for understanding their performance and limitations.

Revolutionizing flame detection: Novelization in flame detection through transferring distillation for knowledge to pruned model

Traditional flame sensors have demonstrated suboptimal detection performance in complex environments, prompting researchers to integrate deep neural network algorithms into these sensors to enhance detection accuracy. However, these algorithms usually rely on sizeable neural networks, resulting in excessive parameterization, which limits their adaptability to different devices. To effectively compress models, this study proposes an effective flame detection framework where the distillation for knowledge is transferred to pruned model (DK-TPM) for the purpose of reducing model parameters. This framework integrates CFasterNet (CFN) and RepConv as foundational components within the base model called YO-CR to accomplish comprehensive feature learning while ensuring better compatibility with devices. Subsequently, structured pruning is applied to the base model to construct a student model. Finally, a new technique called feature distillation with smoothing (FDS) is employed to transfer flame features from the teacher model into the student model. We conducted experiments using three V100 GPUs on five public datasets, demonstrating the generalization capability of the algorithm. Moreover, systematic experiments on flame datasets yielded significant results, achieving an average precision (AP0.5) of 81.5% and frames per second (FPSGPU) of 1048.6, surpassing the requirements of existing devices. Therefore, the proposed DK-TPM is able to provide reliable prediction accuracy and speed for flame detection tasks.

Non-volatile double-tunable vortex metalens design based on Sb2S3 using deep neural network and particle swarm optimization algorithm

A dual-tunable vortex metalens is proposed that leverages deep neural network (DNN) and particle swarm optimization algorithm (PSO), incorporating the use of the phase-change material Sb2S3, which enables simultaneous tunable of both the focal length of the focused vortex beam and the topological charge. The control of traditional vortex metalens relies on changing the polarisation state of the incident light; however, its control range is limited. We employed neural network algorithms to learn the complex mapping relationship between the unit structure and the phase, achieving a fast and accurate prediction of the vortex metalens's phase. Compared with traditional simulation methods, the application of this neural network significantly reduces the time consumption of the design process and avoids the need for re-simulation. Furthermore, our design transcends limitations associated with specific wavelengths and enables the rapid design of tunable metalens across the 1–1.5 μm wavelength range, offering flexibility for the development of dual-tunable metalens. Based on this scheme, we simulated vortex metalens with tunable focal length and topological charge, and the results showed that the simulation results were consistent with the theoretical predictions.

Development and Validation of an Automated Classifier to Diagnose Acute Otitis Media in Children

Acute otitis media (AOM) is a frequently diagnosed illness in children, yet the accuracy of diagnosis has been consistently low. Multiple neural networks have been developed to recognize the presence of AOM with limited clinical application. To develop and internally validate an artificial intelligence decision-support tool to interpret videos of the tympanic membrane and enhance accuracy in the diagnosis of AOM. This diagnostic study analyzed otoscopic videos of the tympanic membrane captured using a smartphone during outpatient clinic visits at 2 sites in Pennsylvania between 2018 and 2023. Eligible participants included children who presented for sick visits or wellness visits. Otoscopic examination. Using the otoscopic videos that were annotated by validated otoscopists, a deep residual-recurrent neural network was trained to predict both features of the tympanic membrane and the diagnosis of AOM vs no AOM. The accuracy of this network was compared with a second network trained using a decision tree approach. A noise quality filter was also trained to prompt users that the video segment acquired may not be adequate for diagnostic purposes. Using 1151 videos from 635 children (majority younger than 3 years of age), the deep residual-recurrent neural network had almost identical diagnostic accuracy as the decision tree network. The finalized deep residual-recurrent neural network algorithm classified tympanic membrane videos into AOM vs no AOM categories with a sensitivity of 93.8% (95% CI, 92.6%-95.0%) and specificity of 93.5% (95% CI, 92.8%-94.3%) and the decision tree model had a sensitivity of 93.7% (95% CI, 92.4%-94.9%) and specificity of 93.3% (92.5%-94.1%). Of the tympanic membrane features outputted, bulging of the TM most closely aligned with the predicted diagnosis; bulging was present in 230 of 230 cases (100%) in which the diagnosis was predicted to be AOM in the test set. These findings suggest that given its high accuracy, the algorithm and medical-grade application that facilitates image acquisition and quality filtering could reasonably be used in primary care or acute care settings to aid with automated diagnosis of AOM and decisions regarding treatment.

Embedded artificial intelligence system using deep learning and raspberrypi for the detection and classification of melanoma

<div lang="EN-US"><span>Melanoma is a kind of skin cancer that originates in melanocytes responsible for producing melanin, it can be a severe and potentially deadly form of cancer because it can metastasize to other regions of the body if not detected and treated early. To facilitate this process, Recently, various computer-assisted low-cost, reliable, and accurate diagnostic systems have been proposed based on artificial intelligence (AI) algorithms, particularly deep learning techniques. This work proposed an innovative and intelligent system that combines the internet of things (IoT) with a Raspberry Pi connected to a camera and a deep learning model based on the deep convolutional neural network (CNN) algorithm for real-time detection and classification of melanoma cancer lesions. The key stages of our model before serializing to the Raspberry Pi: Firstly, the preprocessing part contains data cleaning, data transformation (normalization), and data augmentation to reduce overfitting when training. Then, the deep CNN algorithm is used to extract the features part. Finally, the classification part with applied Sigmoid Activation Function. The experimental results indicate the efficiency of our proposed classification system as we achieved an accuracy rate of 92%, a precision of 91%, a sensitivity of 91%, and an area under the curve- receiver operating characteristics (AUC-ROC) of 0.9133.</span></div>

K-Means-Based DNN Algorithm for a High Accuracy VLP System

In this paper, a positioning algorithm based on the combination of K-means clustering and deep neural networks (DNNs) is first presented for multiple light emitting diodes (LEDs) integrated with visible light positioning (VLP) systems. We extracted the maximum value from the collected optical power of LEDs, utilizing the ratio of each optical power to this maximum optical power as the input training data. The experimental results demonstrate that the proposed algorithm outperformed the conventional DNN algorithm in terms of anti-jamming capability and positioning accuracy. In addition, the positioning accuracy of the proposed system reached a millimeter level, which is the highest experimental VLP accuracy, to the best of our knowledge.

Prediction of bio-oil yield by machine learning model based on 'enhanced data' training

Bio-oil is widely used and has great application potential. With the development of artificial intelligence, machine learning has been gradually applied in the field of biomass, the data augmentation method is a common operation method for training models in the field of computer or data processing. In this study, the concept of 'enhanced data' was proposed for the first time. According to the composition characteristics of biomass raw materials, the sample data were divided into three different sample sets. Based on the light gradient boosting machine (Light GBM) and deep neural network (DNN) algorithm, a prediction model of bio-oil yield was established. Combined with the analysis of partial correlation, the influence of 'enhanced data' on the prediction accuracy of the model was explored. The results showed that the prediction accuracy of the model was improved to a certain extent after adding 'enhanced data'. The Light GBM model was more suitable for predicting bio-oil yield, and the Light GBM _ c model performed best, with R2 of 0.894, MAE of 3.622, and RMSE of 4.445. At the same time, this study also has certain reference significance for other prediction studies in the field of biomass thermochemical conversion.

Solving the spatial extrapolation problem in flood susceptibility using hybrid machine learning, remote sensing, and GIS.

Floods are arguably the most impactful of natural hazards. The increasing magnitude of their effects on the environment, human life, and economic activities calls for improved management of water resources. Flood susceptibility modeling has been used around the world to reduce the damage caused by flooding, although the extrapolation problem still presents a significant challenge. This study develops a machine learning (ML) model utilizing deep neural network (DNN) and optimization algorithms, namely earthworm optimization algorithm (EOA), wildebeest herd optimization (WHO), biogeography-based optimization (BBO), satin bowerbird optimizer (SBO), grasshopper optimization algorithm (GOA), and particle swarm optimization (PSO), to solve the extrapolation problem in the construction of flood susceptibility models. Quang Nam Province was chosen as a case study as it is subject to the significant impact of intense flooding, and Nghe An Province was selected as the region for extrapolation of the flood susceptibility model. Root mean square error (RMSE), receiver operating characteristic (ROC), the area under the ROC curve (AUC), and accuracy (ACC) were applied to assess and compare the fit of each of the models. The results indicated that the models in this study are a good fit in establishing flood susceptibility maps, all with AUC > 0.9. The deep neural network (DNN)-BBO model enjoyed the best results (AUC = 0.99), followed by DNN-WHO (AUC = 0.99), DNN-SBO (AUC = 0.98), DNN-EOA (AUC = 0.96), DNN-GOA (AUC = 0.95), and finally, DNN-PSO (AUC = 0.92). In addition, the models successfully solved the extrapolation problem. These new models can modify their behavior to evaluate flood susceptibility in different regions of the world. The models in this study distribute a first point of reference for debate on the solution to the extrapolation problem, which can support urban planners and other decision-makers in other coastal regions in Vietnam and other countries.

Survival Analysis of Patients with COVID-19 using Deep Neural Network and Random Forrest Techniques

Introduction: The prediction of the survival chance of coronavirus disease 2019 (COVID-19) patients is as important as the early detection of the coronavirus. Since patient mortality, factors may differ by location, this study concentrated on identifying the influential factors and predicting survival for COVID-19 patients using machine learning methods in Fars province, Iran.Material and Methods: The research dataset was extracted in the period January 21, 2020, to September 25, 2020, and contains 25858 hospitalized patients’ records with 51 features. These records were classified into two categories: death (label 1) and survival (label 0). The methodology of this research is CRISP standard. A comparison was made between the efficiency of two deep neural network and random forest algorithms in predicting survival. Modeling steps were done with Python language in the Google Colab environment.Results: Experimental results demonstrated that the deep neural network algorithm had better performance than random forest with accuracy, precision, recall, F-score, and receiver operating characteristic of 97.2%, 100%, 93.54%, 96.66%, and 97.9%, respectively. Based on the results of the random forest model, history of hypertension, chronic neurological disorders, chronic lung diseases, asthma, chronic kidney disease and, heart disease were the most important risk factors related to death.Conclusion: Deployment of our proposed model allows medical professionals to exercise greater caution during the treatment of patients who are most likely to die due to their medical conditions.