Neural Network-based Solution Research Articles

Segmentation of THz holograms for homogenous illumination

This paper investigates the feasibility of applying the hologram segmentation method for homogeneous illumination. Research focuses on improving the uniformity of the illumination obtained from diffractive optical elements in the THz range. The structures are designed with a modified Ping-Pong algorithm and a neural network-based solution. This method allows for the improvement of uniform illumination distribution with the desired shape. Additionally, the phase modulations of the structures are divided into segments, each responsible for imaging at different distances. Various segment combination methods are investigated, differing in shapes, image plane distances, and illumination types. The obtained image intensity maps allow for the identification of the performance of each combination method. Each of the presented structures shows significant improvements in the uniformity of imaged targets compared to the reference Ping-Pong structure. The presented structures were designed for a narrow band case—260 GHz frequency, which corresponds to 1.15 mm wavelength. The application of diffractive structures for homogenization of illumination shows promise. The created structures perform designed beamforming task with variability of intensity improved up to 23% (standard deviation) or 45% (interquartile range) compared with reference structure.

Graph neural network based human detection in videos during occlusion environments

One of the most difficult perceptual problems for many applications is accurately recognizing the human object in a variety of circumstances. This can be difficult due to obstructions, weather, complex backdrops, cast shadows, and occlusions. Occlusion is a challenging open problem where a detector can only perceive a portion of the target human because of obstacles in the surrounding. In this research, an experimental investigation was conducted using the multi object tracking (MOT17) datasets to construct a graph neural network-based solution for the detection of humans in videos while considering the possibility of occlusion. Graph neural network (GNN) is used for the construction of neural solver model for detecting human object in occlusion scenario. The results obtained shows that this proposed method offers a considerable improvement in efficiency in comparison to the ways that have been used in the past. The values obtained for the standard performance metrics are higher than the state-of-the-art methods.

A novel method for response probability density of nonlinear stochastic dynamic systems

AbstractThis paper presents a novel method for analyzing high-dimensional nonlinear stochastic dynamic systems. In particular, we attempt to obtain the solution of the Fokker–Planck–Kolmogorov (FPK) equation governing the response probability density of the system without using the FPK equation directly. The method consists of several important components including the radial basis function neural networks (RBFNN), Feynman–Kac formula and the short-time Gaussian property of the response process. In the area of solving partial differential equations (PDEs) using neural networks, known as physics-informed neural network (PINN), the proposed method presents an effective alternative for obtaining solutions of PDEs without directly dealing with the equation, thus avoids evaluating the derivatives of the equation. This approach has a potential to make the neural network-based solution more efficient and accurate. Several highly challenging examples of nonlinear stochastic systems are presented in the paper to illustrate the effectiveness of the proposed method in comparison to the equation-based RBFNN approach.

Adaptive Deep Fourier Residual method via overlapping domain decomposition

The Deep Fourier Residual (DFR) method is a specific type of variational physics-informed neural network (VPINN). It provides a robust neural network-based solution to partial differential equations (PDEs). The DFR strategy is based on minimizing the dual norm of the weak residual of a PDE, which is equivalent to minimizing the energy norm of the error. To compute the dual norm of the weak residual, the DFR method employs an orthonormal spectral basis of the test space, known for rectangles or cuboids for multiple function spaces.In this work, we extend the DFR method with ideas of traditional domain decomposition (DD). This enables two improvements: (a) to solve problems in more general polygonal domains and (b) to develop an adaptive refinement technique in the test space using a Döfler marking algorithm. In the former case, we retain the desirable equivalence between the employed loss function and the H1 error under non-restrictive assumptions, numerically demonstrating adherence to explicit bounds in the case of the L-shaped domain problem. In the latter, we show how refinement strategies lead to potentially significant improvements against a reference, classical DFR implementation with a test function space of significantly lower dimensionality, allowing us to better approximate singular solutions at a more reasonable computational cost.

Cardiac Auscultation Using Artificial Intelligence on Different Devices

Having to pay to go to a doctor's office and pay for a medical professional to use a stethoscope is costly and inconvenient. A mobile solution that is cheap and uses a medium that is widespread will make diagnoses more accessible. The objective of our research was to assess how feasible and accurate a mobile device solution to cardiac auscultation is, compared to a digital stethoscope. We used a convolutional neural network-based solution, which used the heart sound audio, collected with a digital stethoscope and smartphone, graphed out on a spectrogram for input. We trained two convolutional neural networks, one on the digital stethoscope audio and the other on the smartphone audio. To analyze the outputs, we used the metrics accuracy, recall, precision, and f1 score. We then compared the outputs of the model trained on the digital stethoscope audio versus the model trained on the smartphone audio. The model trained on the smartphone data typically performed 15% worse than the model trained on stethoscope data in terms of accuracy. Based off these results alone, the hardware technology in phones is still not advanced enough to reliably diagnose with machine learning.

Fast rendering of central and peripheral human visual aberrations across the entire visual field with interactive personalization

With the recent progress made in areas such as head-mounted displays and vision-correcting devices, there is a growing interest in fast and personalized algorithms for simulating aberrated human vision. Existing vision-simulating approaches are generally hindered by the lack of personalization, computational cost of rendering, and limited types of supported aberrations. This paper presents a fast vision simulation method with interactive personalization capabilities for simulating arbitrary central and peripheral aberrations of the human eye. First, we describe a novel, neural network-based solution for efficiently estimating the physical structure of the simulated eye and calculating the necessary Zernike aberration coefficients for computing the point-spread functions with varying pupil sizes, focus distances, and incidence angles. Our new approach operates in the sub-second regime and produces highly accurate outputs, facilitating the interactive personalization of vision simulation. Next, we present an improved PSF interpolation method for an existing tiled PSF splatting algorithm for rendering. The proposed algorithm significantly improves the computational performance and memory efficiency of the previous approach, allowing the simulation of peripheral vision with arbitrary visual aberrations in low-latency applications. Following the description of our new techniques, we evaluate their performance characteristics and simulation accuracies on several different eye conditions and test scenarios and compare our results to several previous vision simulation algorithms.

AI-Enabled Framework for Mobile Network Experimentation Leveraging ChatGPT: Case Study of Channel Capacity Calculation for η-µ Fading and Co-Channel Interference

Artificial intelligence has been identified as one of the main driving forces of innovation in state-of-the-art mobile and wireless networks. It has enabled many novel usage scenarios, relying on predictive models for increasing network management efficiency. However, its adoption requires additional efforts, such as mastering the terminology, tools, and newly required steps of data importing and preparation, all of which increase the time required for experimentation. Therefore, we aimed to automate the manual steps as much as possible while reducing the overall cognitive load. In this paper, we explore the potential use of a novel Chat Generative Pre-trained Transformer (ChatGPT) conversational agent together with a model-driven approach relying on the Neo4j graph database in order to aid experimentation and analytics in the case of wireless network planning. As a case study, we present a derivation of the expression for the channel capacity (CC) metric in the case of η-µ multipath fading and η-µ co-channel interference. Moreover, the derived expression is leveraged for quality of service (QoS) estimation within the software simulation environment. ChatGPT, in synergy with a model-driven approach, is used to automate several steps: data importing, generation of graph construction, and machine learning-related Neo4j queries. According to the achieved outcomes, the proposed QoS estimation method, based on the derived CC expression (with precision up to the fifth significant digit), demonstrates satisfactory accuracy (up to 98%) and faster training than the deep neural network-based solution. On the other hand, compared to the manual approach based on our previous work, ChatGPT-based code generation reduces the time required for experimentation by more than 4 times.

Beam wander prediction with recurrent neural networks.

Among the problems that prevent free-space optical communication systems from becoming a truly mainstream technology is beam wander, which is especially important for structured light beams since beam misalignment introduces additional crosstalk at the receiver. The paper suggests a recurrent neural network-based (RNN) solution to predict beam wander in free space optics (FSO). The approach uses past beam center of mass positions to predict future movement, significantly outperforming various prediction types. The proposed approach is demonstrated using under-sampled experimental data over a 260 m link as a worst-case and over-sampled simulated data as a best-case scenario. In addition to conventional Gaussian beams, Hermite- and Laguerre-Gaussian beam wander is also investigated. With a 20 to 40% improvement in error over naive and linear predictions, while predicting multiple samples ahead in typical situations and overall matching or outperforming considered predictions across all studied scenarios, this method could help mitigate turbulence-induced fading and has potential applications in intelligent re-transmits, quality of service, optimized error correction, maximum likelihood-type algorithms, and predictive adaptive optics.

Reinforcement Learning-Based Joint User Scheduling and Link Configuration in Millimeter-Wave Networks

In this paper, we develop algorithms for joint user scheduling and three types of mmWave link configuration: relay selection, codebook optimization, and beam tracking in millimeter wave (mmWave) networks. Our goal is to design an online controller that dynamically schedules users and configures their links to minimize system delay. To solve this complex scheduling problem, we model it as a dynamic decision-making process and develop two reinforcement learning-based solutions. The first solution is based on deep reinforcement learning (DRL), which leverages the proximal policy optimization to train a neural network-based solution. Due to the potential high sample complexity of DRL, we also propose an empirical multi-armed bandit (MAB)-based solution, which decomposes the decision-making process into a sequential of sub-actions and exploits classic maxweight scheduling and Thompson sampling to decide those sub-actions. Our evaluation of the proposed solutions confirms their effectiveness in providing acceptable system delay. It also shows that the DRL-based solution has better delay performance while the MAB-based solution has a faster training process.

On data-driven underwater acoustic direct localization: Design considerations of a deep neural network-based solution

High reverberation, depth-dependent speed of sound and intricate environmental structures are a few of the challenges that make the development of reliable underwater acoustic localization (UAL) methods hard. Since the nontrivial physics of underwater acoustics and the inherent random effects amount to a highly complex statistical model of the measured data, it is natural to consider deep neural networks (NNs) as a central tool in the development of UAL methods. Indeed, recent studies have shown that some properly designed NNs can lead to unprecedented localization capabilities and enhanced accuracy. However, out of the already myriad available possibilities, it is not immediately clear how to choose a proper NN architecture—input structure, layer types, training procedure, loss function(s), etc.—that will lead to successful operation. In this talk, we will present an analytically informed architecture that can learn (for now, in simulations) to localize in complex underwater environments. We will also provide analytical arguments that can explain the robustness of such architectures to certain types of propagation/environment model mismatches. The operation of the discussed architecture will be demonstrated via simulation results in a rich, reverberant channel. The results suggest that deep NNs could become a viable solution to the UAL problem.

Municipal solid waste classification and real-time detection using deep learning methods

Waste management has become a significant issue in most developing countries. Municipal solid waste generation has been steadily increasing over the past decade. Recycling is gaining importance because it is the only method to maintain a healthy and sustainable environment. Furthermore, recycling is not a fully autonomous operation; a significant amount of waste must be performed manually. New and innovative procedures must be implemented to deal with the increasing volume of waste products at recycling facilities. It is suggested that effective solid waste management systems have a practical approach to detecting and classifying waste materials. This study presents CNN and Graph-LSTM, two deep-learning techniques that can recognize typical waste products when handled on a belt conveyor in waste collection systems. This convolutional neural network-based solution is trained to use six object classes: cardboard, metal, glass, plastic, paper, and organic waste. The major advantage is the ability to model long-term dependencies, Improved performance, better generalization, and easy access The experimental findings show that the suggested system can achieve 97.5% accuracy in real-world situations, outperforming existing methods identified in the literature.

AN EFFICIENT ENGLISH TEACHING DRIVEN BY ENTERPRISE-SOCIAL MEDIA BIG DATA: A NEURAL NETWORK-BASED SOLUTION

The cultivation of creativity is closely related to language learning. How to design the creativity promotion mechanism of English teaching in the public environment is the challenge faced by English teachers. With the advent of the era of big data, English teachers can apply the latest research results to classroom teaching. For example, the rational use of social media helps students to learn and communicate in the language, cultivate students’ creativity in learning English, and improves the quality of teaching. Corporate social media has become the most important way for corporate employees to record their lives, express opinions, share and communicate, and it is also one of the reliable and real-time sources of big data that reflects the true state of English learners. Real, accurate, and timely enterprise social media big data samples contain an enormous educational value, providing more possibilities for educational research. From the perspective of value, through sentiment analysis, topic mining, social network analysis, etc. on social media big data, learner portraits can be realized, thereby providing decision-making reference and support for stakeholders. This paper first builds a learning interest classification model based on TCNN-GRU deep learning, collects experimental data sets from an online English learner’s social media platform and performs learning interest classification and labeling, and then uses the TCNN-GRU model to determine the user’s learning interest tendency. On this basis, the concept of learning interest index is further proposed, and a neural network-based corporate social platform English learner portrait model is established. The experimental results show that, compared with the traditional machine learning model, convolutional neural network model, and recurrent neural network model, the TCNN-GRU model can obtain better results in learning interest classification.

Seeing Through Clouds in Satellite Images

This paper presents a neural network-based solution to recover pixels occluded by clouds in satellite images. We leverage radio frequency (RF) signals in the ultra/super-high frequency band that penetrate clouds to help reconstruct the occluded regions in multispectral images. We introduce the first multi-modal multi-temporal method for cloud removal. Our model uses publicly available satellite observations and produces daily cloud-free images. Experimental results show that our system outperforms several baselines on multiple metrics. We also demonstrate use cases of our system in digital agriculture, flood monitoring, and wildfire detection.

A Layer-Wise Surface Deformation Defect Detection by Convolutional Neural Networks in Laser Powder-Bed Fusion Images.

Surface deformation is a multi-factor, laser powder-bed fusion (LPBF) defect that cannot be avoided entirely using current monitoring systems. Distortion and warping, if left unchecked, can compromise the mechanical and physical properties resulting in a build with an undesired geometry. Increasing dwell time, pre-heating the substrate, and selecting appropriate values for the printing parameters are common ways to combat surface deformation. However, the absence of real-time detection and correction of surface deformation is a crucial LPBF problem. In this work, we propose a novel approach to identifying surface deformation problems from powder-bed images in real time by employing a convolutional neural network-based solution. Identifying surface deformation from powder-bed images is a significant step toward real-time monitoring of LPBF. Thirteen bars, with overhangs, were printed to simulate surface deformation defects naturally. The carefully chosen geometric design overcomes problems relating to unlabelled data by providing both normal and defective examples for the model to train. To improve the quality and robustness of the model, we employed several deep learning techniques such as data augmentation and various model evaluation criteria. Our model is 99% accurate in identifying the surface distortion from powder-bed images.

Covid-19 Detection using X-ray

COVID-19 is considered to be the most dangerous and deadly disease for the human body caused by the novel coronavirus. To compact this disease, it is necessary to screen the affected patients in a fast and inexpensive way. With limited testing kits, it is impossible for every patient with respiratory illness to be tested using conventional techniques (RT-PCR). So, Chest X-ray being the most inexpensive and easily available option. Chest X-ray images are primarily used for the diagnosis of this disease. This research has proposed a machine vision approach to detect COVID-19 from the chest X-ray images. The features extracted by the convolutional neural network (CNN) from X-ray images to develop the classification model through training by CNN. Chest X-Ray being the most easily available and least expensive option. In this project, we have proposed a Deep Convolutional Neural Network-based solution which can detect the COVID-19 +ve patients using chest X-Ray images. Multiple state-of-the-art CNN models, have been adopted in the proposed work.

Hall effect on MHD Jeffrey fluid flow with Cattaneo–Christov heat flux model: an application of stochastic neural computing

Exploration and exploitation of intelligent computing infrastructures are becoming of great interest for the research community to investigate different fields of science and engineering offering new improved versions of problem-solving soft computing-based methodologies. The current investigation presents a novel artificial neural network-based solution methodology for the presented problem addressing the properties of Hall current on magneto hydrodynamics (MHD) flow with Jeffery fluid towards a nonlinear stretchable sheet with thickness variation. Generalized heat flux characteristics employing Cattaneo–Christov heat flux model (CCHFM) along with modified Ohms law have been studied. The modelled PDEs are reduced into a dimensionless set of ODEs by introducing appropriate transformations. The temperature and velocity profiles of the fluid are examined numerically with the help of the Adam Bashforth method for different values of physical parameters to study the Hall current with Jeffrey fluid and CCHFM. The examination of the nonlinear input–output with neural network for numerical results is also conducted for the obtained dataset of the system by using Levenberg Marquardt backpropagated networks. The value of Skin friction coefficient, Reynold number, Deborah number, Nusselt number, local wall friction factors and local heat flux are calculated and interpreted for different parameters to have better insight into flow dynamics. The precision level is examined exhaustively by mean square error, error histograms, training states information, regression and fitting plots. Moreover, the performance of the designed solver is certified by mean square error-based learning curves, regression metrics and error histogram analysis. Several significant results for Deborah number, Hall parameters and magnetic field parameters have been presented in graphical and tabular form.

COVD-19 Detection Platform from X-ray Images using Deep Learning

Abstract Since the early days of 2020, COVID-19 has tragic effects on the lives of human beings all over the world. To combat this disease, it is important to survey the infected patients in an inexpensive and fast way. One of the most common ways of achieving this is by performing radiological testing using chest X-Rays and patient coughing sounds. In this work, we propose a Convolutional Neural Network-based solution which is able to identify the positive COVID-19 patients using chest XRay images. Multiple CNN models have been adopted in our work. Each of these models provides a decision whether the patient is affected with COVID-19 or not. Then, a weighted average selection technique is used to provide the final decision. To test the efficiency of our model we have used publicly available chest X-ray images of COVID positive and negative cases. Our approach provided a classification performance of 88.5%. Keywords: COVID-19, CT-Images, Deep Learning, CNN Algorithm.

Load component decomposition and online modelling based on small disturbance response characteristic matching

Accurate load model significantly influences the numerical simulations of power system. The load model obtained by the component-based method does not adequately reflect the load time-varying characteristics, and the measurement-based method relies on fault data, but system faults rarely occur. Inspired by the theory of non-intrusive load monitoring, the load component decomposition and online modelling method based on the small disturbance response characteristic matching is proposed. An optimization model for the load component decomposition based on the curve fitting of small disturbance response characteristics is established. The construction method of a typical load model library is given and a neural network-based solution is designed. The proposed method can determine the load components that constitute the synthesis load model and their proportions by only analysing the small disturbance real-time measurement data. To verify the effectiveness of the proposed method, the distribution of the decomposition errors was analysed by the Monte Carlo simulation. Furthermore, the accuracy of the synthesis load model was verified by a fault simulation using the CEPRI 36-node system. Experimental results show that the proposed method can provide accurate load compositions and dynamically update the synthesis load model for the online power system simulation, planning and operation.

Automatic COVID-19 detection from X-ray images using ensemble learning with convolutional neural network

COVID-19 continues to have catastrophic effects on the lives of human beings throughout the world. To combat this disease it is necessary to screen the affected patients in a fast and inexpensive way. One of the most viable steps towards achieving this goal is through radiological examination, Chest X-Ray being the most easily available and least expensive option. In this paper, we have proposed a Deep Convolutional Neural Network-based solution which can detect the COVID-19 +ve patients using chest X-Ray images. Multiple state-of-the-art CNN models—DenseNet201, Resnet50V2 and Inceptionv3, have been adopted in the proposed work. They have been trained individually to make independent predictions. Then the models are combined, using a new method of weighted average ensembling technique, to predict a class value. To test the efficacy of the solution we have used publicly available chest X-ray images of COVID +ve and –ve cases. 538 images of COVID +ve patients and 468 images of COVID –ve patients have been divided into training, test and validation sets. The proposed approach gave a classification accuracy of 91.62% which is higher than the state-of-the-art CNN models as well the compared benchmark algorithm. We have developed a GUI-based application for public use. This application can be used on any computer by any medical personnel to detect COVID +ve patients using Chest X-Ray images within a few seconds.

Performance Analysis of Deep Neural Network based on Transfer Learning for Pet Classification

Deep learning frameworks have progressed beyond human recognition capabilities and, now it’s the perfect opportunity to optimize them for implementation on the embedded platforms. The present deep learning architectures support learning capabilities, but they lack flexibility for applying learned knowledge on the tasks in other unfamiliar domains. This work tries to fill this gap with the deep neural network-based solution for object detection in unrelated domains with a focus on the reduced footprint of the developed model. Knowledge distillation provides efficient and effective teacher-student learning for a variety of different visual recognition tasks. A lightweight student network can be easily trained under the guidance of the high-capacity teacher networks. The teacher-student architecture implementation on binary classes shows a 20% improvement in accuracy within the same training iterations using the transfer learning approach. The scalability of the student model is tested with binary, ternary and multiclass and their performance is compared on basis of inference speed. The results show that the inference speed does not depend on the number of classes. For similar recognition accuracy, the inference speed of about 50 frames per second or 20ms per image. Thus, this approach can be generalized as per the application requirement with minimal changes, provided the dataset format compatibility.