[go: up one dir, main page]
Next Issue
Volume 16, December
Previous Issue
Volume 16, October
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
4.1
1.8
Journals
Algorithms
Volume 16
Issue 11
share announcement

Algorithms, Volume 16, Issue 11 (November 2023) – 36 articles

Cover Story (view full-size image): In an era where robotics and automation are pivotal, this work offers a crucial insight into computational geometry (CG)—a mathematical discipline for solving geometric problems with practical algorithms. This study traverses the intricate relationship between CG and robotics, particularly in devising paths for mobile robots amidst obstacles. By dissecting CG’s history and methodologies, the paper sheds light on how CG aids in simplifying complex path-planning conundrums. Highlighting the 2021 CG-SHOP competition’s victorious algorithms, the review underscores the real-world impact of CG in enhancing the autonomy of single and multi-robot systems. This comprehensive analysis maps the current terrain of CG applications in robotics and charts a course for future explorations at this cross-disciplinary juncture. View this paper
  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
14 pages, 2161 KiB  
Article
Period Cycle Optimization of Integrated Energy Systems with Long-Term Scheduling Consideration
by Daoyu Ye and Shengxiang Deng
Algorithms 2023, 16(11), 530; https://doi.org/10.3390/a16110530 - 18 Nov 2023
Viewed by 1505
Abstract
The economy and energy saving effects of integrated energy system dispatch plans are influenced by the coupling of different energy devices. In order to consider the impact of changes in equipment load rates on the optimization and scheduling of the system under long-term [...] Read more.
The economy and energy saving effects of integrated energy system dispatch plans are influenced by the coupling of different energy devices. In order to consider the impact of changes in equipment load rates on the optimization and scheduling of the system under long-term operation, a method for energy and component cycle optimization considering energy device capacity and load has been proposed. By improving the initial parameters of the components, energy economic parameters, and operational optimization parameters, the system is subjected to long-term scheduling and multi-cycle operational optimization analysis to evaluate the energy saving and emission reduction potential as well as the economic feasibility of the system. Finally, through numerical analysis, the effectiveness of this optimization approach in achieving energy savings, emission reductions, and cost benefits for the system is validated. Furthermore, compared to existing optimization methods, this approach also assesses the economic feasibility of the system. The case study resulted in a pre-tax IRR of 23.14% and a pre-tax NPV of 66.38 million. It is inferred that the system could generate profits over a 10-year operation period, thereby offering a more rational and cost-effective scheduling scheme for the integrated energy system. Full article
(This article belongs to the Section Algorithms and Mathematical Models for Computer-Assisted Diagnostic Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Schematic representation of the case. (In the figure, yellow lines represent electricity, green represents natural gas, red represents heat load and blue represents cold load).</p> Full article ">Figure 2
<p>Load profiles of various energy components within one operating cycle. ((<b>a</b>) represents the power load curve of each energy equipment in a period; (<b>b</b>) represents the heat load curve of each energy equipment in a period; (<b>c</b>) represents the cold load curve of each energy equipment in a period).</p> Full article ">Figure 3
<p>Energy ratio diagram under three different operating cycles.</p> Full article ">Figure 4
<p>Energy saving and emission reduction curve in each optimization period.</p> Full article ">Figure 5
<p>Cost recovery projection chart.</p> Full article ">
15 pages, 1147 KiB  
Article
A Narrow-Down Approach Based on Machine Learning for Indoor Localization
by Sahibzada Muhammad Ahmad Umair and Tughrul Arslan
Algorithms 2023, 16(11), 529; https://doi.org/10.3390/a16110529 - 17 Nov 2023
Viewed by 1763
Abstract
Over the past decade, the demand and research for indoor localization have burgeoned and Wi-Fi fingerprinting approach has been widely considered because it is cheap and accessible. However, most existing methods lack in terms of positioning accuracy and high computational complexity. To cope [...] Read more.
Over the past decade, the demand and research for indoor localization have burgeoned and Wi-Fi fingerprinting approach has been widely considered because it is cheap and accessible. However, most existing methods lack in terms of positioning accuracy and high computational complexity. To cope with these issues, we formulate a two-stage, coarse and accurate positioning narrow-down approach (NDA). Furthermore, a three-step source domain refinement (SDR) scheme that involves outlier removal, stable AP’s weight enhancement, and a data averaging technique by applying the K-means clustering algorithm is also proposed. The collaboration of SDR scheme with the training data selection, area division, and overlapping schemes reduces the computational complexity and improves coarse positioning accuracy. The effect of the proposed SDR scheme on the performance of the support vector machine (SVM) and random forest algorithms is also presented. In the final/accurate positioning phase, a set of lightweight neural networks (DNNs), trained on different sub-areas, predict the user’s location. This approach significantly increases positioning accuracy while reducing the online computational complexity at the same time. The experimental results show that the proposed approach outperforms the best solutions presented in the literature. Full article
(This article belongs to the Special Issue Artificial Intelligence-Based Algorithms in Wireless Sensor Networks)
►▼ Show Figures

Figure 1

Figure 1
<p>Architecture of the system’s framework.</p> Full article ">Figure 2
<p>Area division.</p> Full article ">Figure 3
<p>RPs selected to train classifier.</p> Full article ">Figure 4
<p>RPs used to train algorithm for regression.</p> Full article ">Figure 5
<p>Library environment; photo taken from [<a href="#B30-algorithms-16-00529" class="html-bibr">30</a>]; dataset available at the Zenodo repository under the open-source MIT license (<a href="https://doi.org/10.3390/data3010003" target="_blank">https://doi.org/10.3390/data3010003</a>, accessed on 15 February 2023).</p> Full article ">Figure 6
<p>Data compression using SDR scheme.</p> Full article ">Figure 7
<p>Impact of data compression on accuracy (<b>a</b>) SVM; (<b>b</b>) RF.</p> Full article ">Figure 8
<p>Impact of SDR scheme on the classification accuracy (<b>a</b>) SVM; (<b>b</b>) RF.</p> Full article ">Figure 9
<p>Classification accuracy of full, reduced, and SDR-refined dataset (<b>a</b>) SVM; (<b>b</b>) RF.</p> Full article ">Figure 10
<p>Comparison of AED with different methods for 15 months.</p> Full article ">Figure 11
<p>CDFs of different methods compared with DNLoc in the library environment.</p> Full article ">
17 pages, 1369 KiB  
Article
Optimal Integrated Single-Framework Algorithm for the Multi-Level School Bus Network Problem
by Amirreza Nickkar and Young-Jae Lee
Algorithms 2023, 16(11), 528; https://doi.org/10.3390/a16110528 - 16 Nov 2023
Viewed by 1398
Abstract
In many states in the United States, school bus fleets are assigned to serve students sequentially at three levels—high school, middle school, and elementary school; however, in past studies, each of these stages in the problem was considered separately. This study introduces a [...] Read more.
In many states in the United States, school bus fleets are assigned to serve students sequentially at three levels—high school, middle school, and elementary school; however, in past studies, each of these stages in the problem was considered separately. This study introduces a novel integrated school bus problem that considers the sequential operation of fleets for all three levels in a unified framework. An example of a hypothetical network was developed and tested to demonstrate the developed algorithm. The algorithm successfully handled the integration of school buses’ optimal route generation while meeting all constraints. The results showed that the routings with the integrated single-framework algorithm can reduce the total costs by 4.5% to 12.4% compared to the routings with the separated level algorithm. Also, it showed that the total costs of the integrated routing framework for different morning and afternoon time windows are 8.28% less than the same routings (identically reversed) for the morning and afternoon time windows. Full article
(This article belongs to the Special Issue Intelligent Optimization for Transportation, Logistics and Vehicle Routing)
►▼ Show Figures

Figure 1

Figure 1
<p>Simplified conceptual operation of school buses in morning trips.</p> Full article ">Figure 2
<p>Simplified conceptual operation of school buses in afternoon trips.</p> Full article ">Figure 3
<p>The locations of the schools in the hypothetical network.</p> Full article ">Figure 4
<p>The middle school (M1) student distribution and routings.</p> Full article ">
24 pages, 3816 KiB  
Article
Search on an NK Landscape with Swarm Intelligence: Limitations and Future Research Opportunities
by Ren-Raw Chen, Cameron D. Miller and Puay Khoon Toh
Algorithms 2023, 16(11), 527; https://doi.org/10.3390/a16110527 - 16 Nov 2023
Viewed by 1512
Abstract
Swarm intelligence has promising applications for firm search and decision-choice problems and is particularly well suited for examining how other firms influence the focal firm’s search. To evaluate search performance, researchers examining firm search through simulation models typically build a performance landscape. The [...] Read more.
Swarm intelligence has promising applications for firm search and decision-choice problems and is particularly well suited for examining how other firms influence the focal firm’s search. To evaluate search performance, researchers examining firm search through simulation models typically build a performance landscape. The NK model is the leading tool used for this purpose in the management science literature. We assess the usefulness of the NK landscape for simulated swarm search. We find that the strength of the swarm model for examining firm search and decision-choice problems—the ability to model the influence of other firms on the focal firm—is limited to the NK landscape. Researchers will need alternative ways to create a performance landscape in order to use our full swarm model in simulations. We also identify multiple opportunities—endogenous landscapes, agent-specific landscapes, incomplete information, and costly movements—that future researchers can include in landscape development to gain the maximum insights from swarm-based firm search simulations. Full article
(This article belongs to the Collection Feature Papers in Algorithms for Multidisciplinary Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>An example of interaction.</p> Full article ">Figure 2
<p>An example of a binary 3-dimensional NK landscape. Note: The battery (<math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mn>1</mn> </msub> </mrow> </semantics></math>) depends on the screen, the CPU (<math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mn>2</mn> </msub> </mrow> </semantics></math>) depends on the battery, and the screen (<math display="inline"><semantics> <mrow> <msub> <mi>x</mi> <mn>3</mn> </msub> </mrow> </semantics></math>) depends on CPU. The fitness values are presented in parentheses (the global peak is 0.6333 at 1, 1, 1).</p> Full article ">Figure 3
<p>NK landscape in 3-D.</p> Full article ">Figure 4
<p>Search on the NK landscape.</p> Full article ">Figure 5
<p>An example of a binary 3-dimensional NK landscape.</p> Full article ">Figure 6
<p>Sorted simulated fitness values (<span class="html-italic">N</span> = 6 and <span class="html-italic">K</span> = 2).</p> Full article ">Figure 7
<p><span class="html-italic">N</span> = 6 and <span class="html-italic">K</span> = 2. Each simulated firm is represented by a different color.</p> Full article ">Figure 8
<p>An example search path.</p> Full article ">Figure 9
<p><span class="html-italic">N</span> = 6 and <span class="html-italic">K</span> = 5. Each simulated firm is represented by a different color.</p> Full article ">Figure 10
<p><span class="html-italic">N</span> = 6, <span class="html-italic">K</span> = 2, one-mutant search (randomly choose between exploitation 75% and exploration 25%). Each simulated firm is represented by a different color.</p> Full article ">Figure 11
<p><span class="html-italic">N</span> = 6, <span class="html-italic">K</span> = 5, one-mutant search. Each simulated firm is represented by a different color.</p> Full article ">Figure 12
<p><span class="html-italic">N</span> = 6, and K is small. Each simulated firm is represented by a different color.</p> Full article ">Figure 13
<p><span class="html-italic">N</span> = 15 and <span class="html-italic">K</span> = 14. Each simulated firm is represented by a different color.</p> Full article ">
11 pages, 356 KiB  
Article
A General Model for Side Information in Neural Networks
by Tameem Adel and Mark Levene
Algorithms 2023, 16(11), 526; https://doi.org/10.3390/a16110526 - 15 Nov 2023
Viewed by 1686
Abstract
We investigate the utility of side information in the context of machine learning and, in particular, in supervised neural networks. Side information can be viewed as expert knowledge, additional to the input, that may come from a knowledge base. Unlike other approaches, our [...] Read more.
We investigate the utility of side information in the context of machine learning and, in particular, in supervised neural networks. Side information can be viewed as expert knowledge, additional to the input, that may come from a knowledge base. Unlike other approaches, our formalism can be used by a machine learning algorithm not only during training but also during testing. Moreover, the proposed approach is flexible as it caters for different formats of side information, and we do not constrain the side information to be fed into the input layer of the network. A formalism is presented based on the difference between the neural network loss without and with side information, stating that it is useful when adding side information reduces the loss during the test phase. As a proof of concept we provide experimental results for two datasets, the MNIST dataset of handwritten digits and the House Price prediction dataset. For the experiments we used feedforward neural networks containing two hidden layers, as well as a softmax output layer. For both datasets, side information is shown to be useful in that it improves the classification accuracy significantly. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Architecture of a neural network with side information. The arrow from additional side information nodes to the hidden layers indicates that there are weighted links between them. The architecture is instantiated once it is specified which links from side information nodes to the network are active, and furthermore, we do not discount the possibility of having active links directly from side information nodes to the output node.</p> Full article ">
17 pages, 4914 KiB  
Article
White Blood Cell Classification: Convolutional Neural Network (CNN) and Vision Transformer (ViT) under Medical Microscope
by Mohamad Abou Ali, Fadi Dornaika and Ignacio Arganda-Carreras
Algorithms 2023, 16(11), 525; https://doi.org/10.3390/a16110525 - 15 Nov 2023
Cited by 7 | Viewed by 3481
Abstract
Deep learning (DL) has made significant advances in computer vision with the advent of vision transformers (ViTs). Unlike convolutional neural networks (CNNs), ViTs use self-attention to extract both local and global features from image data, and then apply residual connections to feed these [...] Read more.
Deep learning (DL) has made significant advances in computer vision with the advent of vision transformers (ViTs). Unlike convolutional neural networks (CNNs), ViTs use self-attention to extract both local and global features from image data, and then apply residual connections to feed these features directly into a fully networked multilayer perceptron head. In hospitals, hematologists prepare peripheral blood smears (PBSs) and read them under a medical microscope to detect abnormalities in blood counts such as leukemia. However, this task is time-consuming and prone to human error. This study investigated the transfer learning process of the Google ViT and ImageNet CNNs to automate the reading of PBSs. The study used two online PBS datasets, PBC and BCCD, and transferred them into balanced datasets to investigate the influence of data amount and noise immunity on both neural networks. The PBC results showed that the Google ViT is an excellent DL neural solution for data scarcity. The BCCD results showed that the Google ViT is superior to ImageNet CNNs in dealing with unclean, noisy image data because it is able to extract both global and local features and use residual connections, despite the additional time and computational overhead. Full article
(This article belongs to the Special Issue AI Algorithms in Medical Imaging)
►▼ Show Figures

Figure 1

Figure 1
<p>Methodology workflow using the PBC dataset.</p> Full article ">Figure 2
<p>Methodology workflow using the BCCD dataset.</p> Full article ">Figure 3
<p>Eosinophil sample images taken from the PBC and BCCD datasets.</p> Full article ">Figure 4
<p>Architecture of VGG-16 model classifying a neutrophil.</p> Full article ">Figure 5
<p>PBC neutrophil image resizing and splitting.</p> Full article ">Figure 6
<p>Architecture of ViT classifying a neutrophil.</p> Full article ">Figure 7
<p>Underfitting and overfitting.</p> Full article ">Figure 8
<p>Score-CAM for DenseNet-169 model fitted with the DS-3 PBC dataset.</p> Full article ">
17 pages, 8845 KiB  
Article
Utilizing Mixture Regression Models for Clustering Time-Series Energy Consumption of a Plastic Injection Molding Process
by Massimo Pacella, Matteo Mangini and Gabriele Papadia
Algorithms 2023, 16(11), 524; https://doi.org/10.3390/a16110524 - 15 Nov 2023
Cited by 1 | Viewed by 1832
Abstract
Considering the issue of energy consumption reduction in industrial plants, we investigated a clustering method for mining the time-series data related to energy consumption. The industrial case study considered in our work is one of the most energy-intensive processes in the plastics industry: [...] Read more.
Considering the issue of energy consumption reduction in industrial plants, we investigated a clustering method for mining the time-series data related to energy consumption. The industrial case study considered in our work is one of the most energy-intensive processes in the plastics industry: the plastic injection molding process. Concerning the industrial setting, the energy consumption of the injection molding machine was monitored across multiple injection molding cycles. The collected data were then analyzed to establish patterns and trends in the energy consumption of the injection molding process. To this end, we considered mixtures of regression models given their flexibility in modeling heterogeneous time series and clustering time series in an unsupervised machine learning framework. Given the assumption of autocorrelated data and exogenous variables in the mixture model, we implemented an algorithm for model fitting that combined autocorrelated observations with spline and polynomial regressions. Our results demonstrate an accurate grouping of energy-consumption profiles, where each cluster is related to a specific production schedule. The clustering method also provides a unique profile of energy consumption for each cluster, depending on the production schedule and regression approach (i.e., spline and polynomial). According to these profiles, information related to the shape of energy consumption was identified, providing insights into reducing the electrical demand of the plant. Full article
(This article belongs to the Special Issue Algorithms in Data Classification)
►▼ Show Figures

Figure 1

Figure 1
<p>Process flow chart for fins production. The production of fins is managed through two IIMs (machine 1 and 2). Machine 1 is used for the creation of the base product, namely paddles, while machine 2 is used for colored booties. (Image courtesy of STAM and SEACSUB).</p> Full article ">Figure 2
<p>Historical time-series dataset comprising 30 instances of hourly energy consumption in kilowatt-hours (kWh) on machine 1 on 24 h for 30 days. The vertical axis represents the energy consumption in kWh, and a different color is used to distinguish between the various production days in the dataset.</p> Full article ">Figure 3
<p>Collection and recording of hourly energy consumption data of nine IMMs into the SQL database for effective data management and analysis.</p> Full article ">Figure 4
<p>Graphical representation of the centroids for the two clusters generated through the implementation of the <span class="html-italic">K</span>-means algorithm with <math display="inline"><semantics> <mrow> <mi>K</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>. Varying colors differentiate profiles.</p> Full article ">Figure 5
<p>Graphical representation of the centroids for the two clusters generated through the implementation of the SC algorithm with <math display="inline"><semantics> <mrow> <mi>K</mi> <mo>=</mo> <mn>3</mn> </mrow> </semantics></math>. Varying colors differentiate profiles.</p> Full article ">Figure 6
<p>Graphical depiction of the centroids for each of the six clusters (1–6) resulting from the implementation of B-spline regression mixtures. Varying colors differentiate profiles.</p> Full article ">Figure 7
<p>Graphical representation of the B-spline of order 3 resulting from the EM algorithm resampled at a frequency of a minute in 24 h (1440 data points).</p> Full article ">Figure 8
<p>Sixteen energy profiles grouped into Cluster 2, exhibiting a high-load profile spanning a period of approximately 16 h. Varying colors differentiate profiles.</p> Full article ">Figure 9
<p>Six energy profiles grouped into Cluster 3, exhibiting a high-load profile spanning a period of approximately 7 h during the first half of the day. Varying colors differentiate profiles.</p> Full article ">Figure 10
<p>Five energy profiles grouped into Cluster 6, exhibiting a high-load profile spanning a period of approximately 7 h during the second half of the day. Varying colors differentiate profiles.</p> Full article ">Figure 11
<p>Graphical depiction of the centroids for each of the six clusters (1–6) resulting from the implementation of polynomial regression mixtures. Varying colors differentiate profiles.</p> Full article ">
15 pages, 936 KiB  
Article
An Intelligent Injury Rehabilitation Guidance System for Recreational Runners Using Data Mining Algorithms
by Theodoros Tzelepis, George Matlis, Nikos Dimokas, Petros Karvelis, Paraskevi Malliou and Anastasia Beneka
Algorithms 2023, 16(11), 523; https://doi.org/10.3390/a16110523 - 15 Nov 2023
Cited by 2 | Viewed by 1443
Abstract
In recent years the number of people who exercise every day has increased dramatically. More precisely, due to COVID period many people have become recreational runners. Recreational running is a regular way to keep active and healthy at any age. Additionally, running is [...] Read more.
In recent years the number of people who exercise every day has increased dramatically. More precisely, due to COVID period many people have become recreational runners. Recreational running is a regular way to keep active and healthy at any age. Additionally, running is a popular physical exercise that offers numerous health advantages. However, recreational runners report a high incidence of musculoskeletal injuries due to running. The healthcare industry has been compelled to use information technology due to the quick rate of growth and developments in electronic systems, the internet, and telecommunications. Our proposed intelligent system uses data mining algorithms for the rehabilitation guidance of recreational runners with musculoskeletal discomfort. The system classifies recreational runners based on a questionnaire that has been built according to the severity, irritability, nature, stage, and stability model and advise them on the appropriate treatment plan/exercises to follow. Through rigorous testing across various case studies, our method has yielded highly promising results, underscoring its potential to significantly contribute to the well-being and rehabilitation of recreational runners facing musculoskeletal challenges. Full article
(This article belongs to the Special Issue Algorithms in Data Classification)
►▼ Show Figures

Figure 1

Figure 1
<p>A SMOTE example in a 2D feature space. <span class="html-italic">A</span>1 is the randomly selected sample from the minority class, and <span class="html-italic">A</span>2 is its nearest neighbor. <span class="html-italic">S</span>1 is the generated synthetic sample along the line segment of their calculated distance.</p> Full article ">Figure 2
<p>Decision tree classifier for a 30% test set splitting. The gini value [<a href="#B34-algorithms-16-00523" class="html-bibr">34</a>] at each node evaluates how well a decision tree node separates the data into different categories. Its values range from 0 to 1. A value of 0 indicates a perfect separation, meaning all the samples at one node belong to one category. A value of 1 indicates an even distribution of samples among the different categories.</p> Full article ">Figure 3
<p>The intelligent system architecture.</p> Full article ">Figure 4
<p>The questionnaire web page.</p> Full article ">Figure 5
<p>The results web page.</p> Full article ">
22 pages, 2044 KiB  
Article
Relational Fisher Analysis: Dimensionality Reduction in Relational Data with Global Convergence
by Li-Na Wang, Guoqiang Zhong, Yaxin Shi and Mohamed Cheriet
Algorithms 2023, 16(11), 522; https://doi.org/10.3390/a16110522 - 15 Nov 2023
Viewed by 1697
Abstract
Most of the dimensionality reduction algorithms assume that data are independent and identically distributed (i.i.d.). In real-world applications, however, sometimes there exist relationships between data. Some relational learning methods have been proposed, but those with discriminative relationship analysis are lacking yet, as important [...] Read more.
Most of the dimensionality reduction algorithms assume that data are independent and identically distributed (i.i.d.). In real-world applications, however, sometimes there exist relationships between data. Some relational learning methods have been proposed, but those with discriminative relationship analysis are lacking yet, as important supervisory information is usually ignored. In this paper, we propose a novel and general framework, called relational Fisher analysis (RFA), which successfully integrates relational information into the dimensionality reduction model. For nonlinear data representation learning, we adopt the kernel trick to RFA and propose the kernelized RFA (KRFA). In addition, the convergence of the RFA optimization algorithm is proved theoretically. By leveraging suitable strategies to construct the relational matrix, we conduct extensive experiments to demonstrate the superiority of our RFA and KRFA methods over related approaches. Full article
(This article belongs to the Special Issue Machine Learning for Pattern Recognition)
►▼ Show Figures

Figure 1

Figure 1
<p>One image of the Ibn Sina dataset.</p> Full article ">Figure 2
<p>Classification accuracy obtained by LDA, MFA and RFA on the Ibn Sina dataset.</p> Full article ">Figure 3
<p>Sample images from the USPS dataset.</p> Full article ">Figure 4
<p>Classification results obtained by RFA and the compared methods on the USPS dataset.</p> Full article ">Figure 5
<p>3D visualization of the mapped data obtained by RFA. Samples of different classes are marked with different colors.</p> Full article ">Figure 6
<p>2D visualization for data pairs. (<b>a</b>–<b>d</b>) are the results obtained by RFA, and (<b>e</b>–<b>h</b>) are the results obtained by MFA.</p> Full article ">Figure 7
<p>Classification results obtained by RFA and MFA with different values of <math display="inline"><semantics> <msub> <mi>k</mi> <mi>i</mi> </msub> </semantics></math> on the USPS dataset.</p> Full article ">Figure 8
<p>Classification results obtained by RFA and MFA with different values of <math display="inline"><semantics> <msub> <mi>k</mi> <mi>p</mi> </msub> </semantics></math> on the USPS dataset.</p> Full article ">Figure 9
<p>Changing cave of <math display="inline"><semantics> <mi>η</mi> </semantics></math> over the iteration number on the PIE dataset.</p> Full article ">Figure 10
<p>Changing cave of <math display="inline"><semantics> <mi>η</mi> </semantics></math> over the iteration number on the YaleB dataset.</p> Full article ">Figure 11
<p>Classification accuracy and standard deviation obtained by RFA and PRPCA on the document and webpage classification problems.</p> Full article ">
14 pages, 4300 KiB  
Article
Automatic Multiorgan Segmentation in Pelvic Region with Convolutional Neural Networks on 0.35 T MR-Linac Images
by Emmanouil Koutoulakis, Louis Marage, Emmanouil Markodimitrakis, Leone Aubignac, Catherine Jenny, Igor Bessieres and Alain Lalande
Algorithms 2023, 16(11), 521; https://doi.org/10.3390/a16110521 - 15 Nov 2023
Viewed by 1787
Abstract
MR-Linac is a recent device combining a linear accelerator with an MRI scanner. The improved soft tissue contrast of MR images is used for optimum delineation of tumors or organs at risk (OARs) and precise treatment delivery. Automatic segmentation of OARs can contribute [...] Read more.
MR-Linac is a recent device combining a linear accelerator with an MRI scanner. The improved soft tissue contrast of MR images is used for optimum delineation of tumors or organs at risk (OARs) and precise treatment delivery. Automatic segmentation of OARs can contribute to alleviating the time-consuming process for radiation oncologists and improving the accuracy of radiation delivery by providing faster, more consistent, and more accurate delineation of target structures and organs at risk. It can also help reduce inter-observer variability and improve the consistency of contouring while reducing the time required for treatment planning. In this work, state-of-the-art deep learning techniques were evaluated based on 2D and 2.5D training strategies to develop a comprehensive tool for the accurate segmentation of pelvic OARs dedicated to 0.35 T MR-Linac. In total, 103 cases with 0.35 T MR images of the pelvic region were investigated. Experts considered and contoured the bladder, rectum, and femoral heads as OARs and the prostate as the target volume. For the training of the neural network, 85 patients were randomly selected, and 18 were used for testing. Multiple U-Net-based architectures were considered, and the best model was compared using both 2D and 2.5D training strategies. The evaluation of the models was performed based on two metrics: the Dice similarity coefficient (DSC) and the Hausdorff distance (HD). In the 2D training strategy, Residual Attention U-Net (ResAttU-Net) had the highest scores among the other deep neural networks. Due to the additional contextual information, the configured 2.5D ResAttU-Net performed better. The overall DSC were 0.88 ± 0.09 and 0.86 ± 0.10, and the overall HD was 1.78 ± 3.02 mm and 5.90 ± 7.58 mm for 2.5D and 2D ResAttU-Net, respectively. The 2.5D ResAttU-Net provides accurate segmentation of OARs without affecting the computational cost. The developed end-to-end pipeline will be merged with the treatment planning system for in-time automatic segmentation. Full article
(This article belongs to the Special Issue Artificial Intelligence for Medical Imaging)
►▼ Show Figures

Figure 1

Figure 1
<p>Data sample. (<b>a</b>) MR image in axial orientation, and (<b>b</b>) ground truth segmented OARs and prostate from expert physicians. Each color on (<b>b</b>) corresponds to a different organ.</p> Full article ">Figure 2
<p>Minimum filtering: (<b>a</b>) the pixels with a normalized gray level &gt;0.1, and (<b>b</b>) the results after the minimum filtering, removing background noise.</p> Full article ">Figure 3
<p>Residual Attention U-Net model. The residual blocks were replaced with the plain convolutional layers. The left side of the model represents the encoding path, and the right side represents the decoding path, respectively.</p> Full article ">Figure 4
<p>(<b>a</b>) The attention block, where W<sub>g</sub> is the feature map derived from the encoding path, and W<sub>x</sub> and X are the convolved and normalized output of the previous residual block. (<b>b</b>) The structure of the residual block.</p> Full article ">Figure 5
<p>Automatic segmentation and post-processing. (<b>a</b>) The predicted mask which contains false-predicted pixels. (<b>b</b>) Final mask after the post-processing steps, where the wrong segment of the rectum is removed, and the bladder’s segmentation morphology was corrected.</p> Full article ">Figure 6
<p>Visualization of the splits regarding the 5-fold cross-validation.</p> Full article ">Figure 7
<p>Sample visualization in 3D of a predicted volume.</p> Full article ">Figure 8
<p>Segmentation using 2.5D ResAttU-Net. The manual delineation is highlighted with red and the predicted of our model with green.</p> Full article ">
16 pages, 3320 KiB  
Article
Improved Object Detection Method Utilizing YOLOv7-Tiny for Unmanned Aerial Vehicle Photographic Imagery
by Linhua Zhang, Ning Xiong, Xinghao Pan, Xiaodong Yue, Peng Wu and Caiping Guo
Algorithms 2023, 16(11), 520; https://doi.org/10.3390/a16110520 - 14 Nov 2023
Cited by 14 | Viewed by 4851
Abstract
In unmanned aerial vehicle photographs, object detection algorithms encounter challenges in enhancing both speed and accuracy for objects of different sizes, primarily due to complex backgrounds and small objects. This study introduces the PDWT-YOLO algorithm, based on the YOLOv7-tiny model, to improve the [...] Read more.
In unmanned aerial vehicle photographs, object detection algorithms encounter challenges in enhancing both speed and accuracy for objects of different sizes, primarily due to complex backgrounds and small objects. This study introduces the PDWT-YOLO algorithm, based on the YOLOv7-tiny model, to improve the effectiveness of object detection across all sizes. The proposed method enhances the detection of small objects by incorporating a dedicated small-object detection layer, while reducing the conflict between classification and regression tasks through the replacement of the YOLOv7-tiny model’s detection head (IDetect) with a decoupled head. Moreover, network convergence is accelerated, and regression accuracy is improved by replacing the Complete Intersection over Union (CIoU) loss function with a Wise Intersection over Union (WIoU) focusing mechanism in the loss function. To assess the proposed model’s effectiveness, it was trained and tested on the VisDrone-2019 dataset comprising images captured by various drones across diverse scenarios, weather conditions, and lighting conditions. The experiments show that [email protected]:0.95 and [email protected] increased by 5% and 6.7%, respectively, with acceptable running speed compared with the original YOLOv7-tiny model. Furthermore, this method shows improvement over other datasets, confirming that PDWT-YOLO is effective for multiscale object detection. Full article
(This article belongs to the Section Evolutionary Algorithms and Machine Learning)
►▼ Show Figures

Figure 1

Figure 1
<p>YOLOv7-tiny network structure.</p> Full article ">Figure 2
<p>PDWT-YOLO network structure.</p> Full article ">Figure 3
<p>Coupled and decoupled heads compared.</p> Full article ">Figure 4
<p>Sample images from VisDrone-2019 dataset.</p> Full article ">Figure 5
<p>Label distribution of dataset.</p> Full article ">Figure 6
<p>Dataset detection effect comparison. (<b>a</b>) YOLOv7-tiny, and (<b>b</b>) PDWT-YOLO.</p> Full article ">Figure 7
<p>Training comparison of YOLOv7-tiny after adding a decoupled head (B1), P2 layer small-object detection head (B2), and WIoU (PDWT-YOLO). (<b>a</b>) mAP@0.5, and (<b>b</b>) mAP@0.5:0.95.</p> Full article ">
6 pages, 216 KiB  
Communication
Two Kadane Algorithms for the Maximum Sum Subarray Problem
by Joseph B. Kadane
Algorithms 2023, 16(11), 519; https://doi.org/10.3390/a16110519 - 14 Nov 2023
Cited by 1 | Viewed by 2004
Abstract
The maximum sum subarray problem is to find a contiguous subarray with the largest sum. The history of algorithms to address this problem is recounted, culminating in what is known as Kadane’s algorithm. However, that algorithm is not the algorithm Kadane intended. Nonetheless, [...] Read more.
The maximum sum subarray problem is to find a contiguous subarray with the largest sum. The history of algorithms to address this problem is recounted, culminating in what is known as Kadane’s algorithm. However, that algorithm is not the algorithm Kadane intended. Nonetheless, the algorithm known as Kadane’s has found many uses, some of which are recounted here. The algorithm Kadane intended is reported here, and compared to the algorithm attributed to Kadane. They are both linear in time, employ just a few words of memory, and use a dynamic programming structure. The results proved here show that these two algorithms differ only in the case of an input consisting of only negative numbers. In that case, the algorithm Kadane intended is more informative than the algorithm attributed to him. Full article
(This article belongs to the Section Combinatorial Optimization, Graph, and Network Algorithms)
18 pages, 2282 KiB  
Review
Performance and Applicability of Post-Quantum Digital Signature Algorithms in Resource-Constrained Environments
by Marin Vidaković and Kruno Miličević
Algorithms 2023, 16(11), 518; https://doi.org/10.3390/a16110518 - 13 Nov 2023
Cited by 3 | Viewed by 3406
Abstract
The continuous development of quantum computing necessitates the development of quantum-resistant cryptographic algorithms. In response to this demand, the National Institute of Standards and Technology selected standardized algorithms including Crystals-Dilithium, Falcon, and Sphincs+ for digital signatures. This paper provides a comparative evaluation of [...] Read more.
The continuous development of quantum computing necessitates the development of quantum-resistant cryptographic algorithms. In response to this demand, the National Institute of Standards and Technology selected standardized algorithms including Crystals-Dilithium, Falcon, and Sphincs+ for digital signatures. This paper provides a comparative evaluation of these algorithms across key metrics. The results indicate varying strengths and weaknesses for each algorithm, underscoring the importance of context-specific deployments. Our findings indicate that Dilithium offers advantages in low-power scenarios, Falcon excels in signature verification speed, and Sphincs+ provides robust security at the cost of computational efficiency. These results underscore the importance of context-specific deployments in specific and resource-constrained technological applications, like IoT, smart cards, blockchain, and vehicle-to-vehicle communication. Full article
(This article belongs to the Special Issue Surveys in Algorithm Analysis and Complexity Theory, Part II)
►▼ Show Figures

Figure 1

Figure 1
<p>Execution speed of Dilithium (cycle counts) [<a href="#B19-algorithms-16-00518" class="html-bibr">19</a>].</p> Full article ">Figure 2
<p>Comparison of Dilithium and Falcon by public key and signature lengths (Falcon’s values for Level 3 are not available) [<a href="#B20-algorithms-16-00518" class="html-bibr">20</a>].</p> Full article ">Figure 3
<p>Key and signature size for Falcon and some competing algorithms [<a href="#B24-algorithms-16-00518" class="html-bibr">24</a>].</p> Full article ">Figure 4
<p>Runtime benchmarks for Sphincs+, Sphincs+-Haraka on AES-NI, Sphincs+-Sha-256 on AVX2, Sphincs+-Shake256 on AVX2 [<a href="#B25-algorithms-16-00518" class="html-bibr">25</a>].</p> Full article ">Figure 5
<p>Comparison of public keys (left bars/dots) and signatures (right bars/dots) according to [<a href="#B19-algorithms-16-00518" class="html-bibr">19</a>,<a href="#B23-algorithms-16-00518" class="html-bibr">23</a>,<a href="#B26-algorithms-16-00518" class="html-bibr">26</a>] (bars) and [<a href="#B27-algorithms-16-00518" class="html-bibr">27</a>] (dots).</p> Full article ">Figure 6
<p>Execution speeds (in thousands of clock cycles) of digital signature algorithms on the Arm Cortex M4 platform according to [<a href="#B37-algorithms-16-00518" class="html-bibr">37</a>] (“Dilithium 1”, “Falcon 1” and “Sphincs+” bars), [<a href="#B9-algorithms-16-00518" class="html-bibr">9</a>] (“Falcon 2” bars) and [<a href="#B21-algorithms-16-00518" class="html-bibr">21</a>] (“Dilithium 2” bars).</p> Full article ">Figure 7
<p>Key generation speeds (in milliseconds) for digital signature algorithms for the i7-6700 processor (Dilithium for NIST levels 1–3, other algorithms for levels 1–5; bars) [<a href="#B38-algorithms-16-00518" class="html-bibr">38</a>] and the i7-1165G7 processor (Dilithium 3 and Falcon 512; blue and green dots) [<a href="#B27-algorithms-16-00518" class="html-bibr">27</a>].</p> Full article ">
20 pages, 5506 KiB  
Article
Comparison of Machine Learning Classifiers for the Detection of Breast Cancer in an Electrical Impedance Tomography Setup
by Jöran Rixen, Nico Blass, Simon Lyra and Steffen Leonhardt
Algorithms 2023, 16(11), 517; https://doi.org/10.3390/a16110517 - 13 Nov 2023
Cited by 1 | Viewed by 1910
Abstract
Breast cancer is the leading cause of cancer-related death among women. Early prediction is crucial as it severely increases the survival rate. Although classical X-ray mammography is an established technique for screening, many eligible women do not consider this due to concerns about [...] Read more.
Breast cancer is the leading cause of cancer-related death among women. Early prediction is crucial as it severely increases the survival rate. Although classical X-ray mammography is an established technique for screening, many eligible women do not consider this due to concerns about pain from breast compression. Electrical Impedance Tomography (EIT) is a technique that aims to visualize the conductivity distribution within the human body. As cancer has a greater conductivity than surrounding fatty tissue, it provides a contrast for image reconstruction. However, the interpretation of EIT images is still hard, due to the low spatial resolution. In this paper, we investigated three different classification models for the detection of breast cancer. This is important as EIT is a highly non-linear inverse problem and tends to produce reconstruction artifacts, which can be misinterpreted as, e.g., tumors. To aid in the interpretation of breast cancer EIT images, we compare three different classification models for breast cancer. We found that random forests and support vector machines performed best for this task. Full article
(This article belongs to the Special Issue Artificial Intelligence for Medical Imaging)
►▼ Show Figures

Figure 1

Figure 1
<p>Structure of the female breast.</p> Full article ">Figure 2
<p>Visualization of the used FEM model with the electrodes placed in green. The values on the axes are given in <math display="inline"><semantics> <mrow> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math>.</p> Full article ">Figure 3
<p>Visualization of the glandular tissue structure with a tumor visible in yellow.</p> Full article ">Figure 4
<p>Comparison of the F1-Score between SVM and RF for noiseless data with tumor radii from <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>12</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math>.</p> Full article ">Figure 5
<p>Influence of noise for different tumor radii on the SVM.</p> Full article ">Figure 6
<p>Influence of noise for different tumor radii on the RF.</p> Full article ">Figure 7
<p>Performance of the SVM for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure 8
<p>Performance of the RF for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure 9
<p>Performance of the ANN for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure 10
<p>Performance of the SVM for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure 11
<p>Performance of the RF for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure 12
<p>Performance of the ANN for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure 13
<p>Performance of the SVM for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure 14
<p>Performance of the RF for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure 15
<p>Performance of the ANN for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure A1
<p>Comparison of the Sensitivity between SVM and RF for noiseless data with tumor radii from <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>12</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math>.</p> Full article ">Figure A2
<p>Comparison of the Precision between SVM and RF for noiseless data with tumor radii from <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>12</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math>.</p> Full article ">Figure A3
<p>Influence of noise for different tumor radii on the SVM’s Sensitivity.</p> Full article ">Figure A4
<p>Influence of noise for different tumor radii on the SVM’s Precision.</p> Full article ">Figure A5
<p>Influence of noise for different tumor radii on the RF’s Sensitivity.</p> Full article ">Figure A6
<p>Influence of noise for different tumor radii on the RF’s Precision.</p> Full article ">Figure A7
<p>Performance of the SVM’s Sensitivity for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure A8
<p>Performance of the SVM’s Precision for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure A9
<p>Performance of the RF’s Sensitivity for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure A10
<p>Performance of the RF’s Precision for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure A11
<p>Performance of the ANN’s Sensitivity for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure A12
<p>Performance of the ANN’s Precision for fixed noise levels on different test data. The blue circular markers indicate a match in noise level between test and training data.</p> Full article ">Figure A13
<p>Performance of the SVM’s Sensitivity for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure A14
<p>Performance of the SVM’s Precision for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure A15
<p>Performance of the RF’s Sensitivity for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure A16
<p>Performance of the RF’s Precision for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure A17
<p>Performance of the ANN’s Sensitivity for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure A18
<p>Performance of the ANN’s Precision for a fixed tumor radii ranges, while testing was done on the whole range of tumor radii.</p> Full article ">Figure A19
<p>Performance of the SVM’s Sensitivity for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure A20
<p>Performance of the SVM’s Precision for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure A21
<p>Performance of the RF’s Sensitivity for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure A22
<p>Performance of the RF’s Precision for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure A23
<p>Performance of the ANN’s Sensitivity for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">Figure A24
<p>Performance of the ANN’s Precision for a tumor radius range of <math display="inline"><semantics> <mrow> <mn>6</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> to <math display="inline"><semantics> <mrow> <mn>13</mn> <mspace width="0.166667em"/> <mi mathvariant="normal">m</mi> <mi mathvariant="normal">m</mi> </mrow> </semantics></math> and an SNR of 30, while testing was done different radii and SNR levels.</p> Full article ">
26 pages, 24948 KiB  
Article
Research on a Classification Method for Strip Steel Surface Defects Based on Knowledge Distillation and a Self-Adaptive Residual Shrinkage Network
by Xinbo Huang, Zhiwei Song, Chao Ji, Ye Zhang and Luya Yang
Algorithms 2023, 16(11), 516; https://doi.org/10.3390/a16110516 - 10 Nov 2023
Cited by 1 | Viewed by 1498
Abstract
Different types of surface defects will occur during the production of strip steel. To ensure production quality, it is essential to classify these defects. Our research indicates that two main problems exist in the existing strip steel surface defect classification methods: (1) they [...] Read more.
Different types of surface defects will occur during the production of strip steel. To ensure production quality, it is essential to classify these defects. Our research indicates that two main problems exist in the existing strip steel surface defect classification methods: (1) they cannot solve the problem of unbalanced data using few-shot in reality, (2) they cannot meet the requirement of online real-time classification. To solve the aforementioned problems, a relational knowledge distillation self-adaptive residual shrinkage network (RKD-SARSN) is presented in this work. First, the data enhancement strategy of Cycle GAN defective sample migration is designed. Second, the self-adaptive residual shrinkage network (SARSN) is intended as the backbone network for feature extraction. An adaptive loss function based on accuracy and geometric mean (Gmean) is proposed to solve the problem of unbalanced samples. Finally, a relational knowledge distillation model (RKD) is proposed, and the functions of GUI operation interface encapsulation are designed by combining image processing technology. SARSN is used as a teacher model, its generalization performance is transferred to the lightweight network ResNet34, and it is conveniently deployed as a student model. The results show that the proposed method can improve the deployment efficiency of the model and ensure the real-time performance of the classification algorithms. It is superior to other mainstream algorithms for fine-grained images with unbalanced data classification. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>(<b>a</b>) Model deployment process. (<b>b</b>) The framework of the proposed method.</p> Full article ">Figure 2
<p>Defect sample migration algorithms.</p> Full article ">Figure 3
<p>Cycle GAN structure diagram.</p> Full article ">Figure 4
<p>Schematic diagram of the residual learning unit. (<b>a</b>) Building block, (<b>b</b>) bottleneck, (<b>c</b>) architecture of a Res Net, and (<b>d</b>) architecture of RSBU.</p> Full article ">Figure 5
<p>The architecture of SARSN.</p> Full article ">Figure 6
<p>Schematic diagram of the SARSN backbone network.</p> Full article ">Figure 7
<p>Relational knowledge distillation.</p> Full article ">Figure 8
<p>Image mosaicing. (<b>a</b>) The image to be spliced. (<b>b</b>) Image mosaicing. (<b>c</b>) Median fuzzy image.</p> Full article ">Figure 9
<p>Image fusion.</p> Full article ">Figure 10
<p>Cycle GAN data migration results.</p> Full article ">Figure 11
<p>SARSN visualization of training process. (<b>a</b>) Crack. (<b>b</b>) Inclusion. (<b>c</b>) Patches. (<b>d</b>) Pitted surface. (<b>e</b>) Rolled-in scale. (<b>f</b>) Scratches.</p> Full article ">Figure 12
<p>Teacher model training loss and accuracy curve. (<b>a</b>) Teacher model training accuracy curve. (<b>b</b>) Teacher model training loss curve.</p> Full article ">Figure 13
<p>Student model respective testing accuracy curve. (<b>a</b>) Original ResNet34 accuracy of each part, (<b>b</b>) Teacher-SARSN50 Student-ResNet34 accuracy of each part.</p> Full article ">Figure 14
<p>The confusion matrix.</p> Full article ">Figure 15
<p>Comparison of classification performance of ResNet, SE-ResNet, and SARSN (<b>a</b>) 34 series, (<b>b</b>) 50 series, (<b>c</b>) 101 series.</p> Full article ">Figure 16
<p>Accuracy curve of the classification test for each model.</p> Full article ">Figure 17
<p>Comparison of the results of each classification algorithm.</p> Full article ">Figure 18
<p>Image processing flow. (<b>a</b>) Original images. (<b>b</b>) Image preprocessing. (<b>c</b>) Boundary tracking algorithm. (<b>d</b>) Connected-domain calibration. (<b>e</b>) ROI extraction. (<b>f</b>) Centroid localization.</p> Full article ">Figure 19
<p>Feature extraction of salient region heat map visualization results. (<b>a</b>) Original image. (<b>b</b>) Visualization results of the salient region.</p> Full article ">Figure 20
<p>The evaluation system for surface defects of strip steel.</p> Full article ">
18 pages, 1378 KiB  
Article
Using Graph Neural Networks for Social Recommendations
by Dharahas Tallapally, John Wang, Katerina Potika and Magdalini Eirinaki
Algorithms 2023, 16(11), 515; https://doi.org/10.3390/a16110515 - 10 Nov 2023
Cited by 1 | Viewed by 2989
Abstract
Recommender systems have revolutionized the way users discover and engage with content. Moving beyond the collaborative filtering approach, most modern recommender systems leverage additional sources of information, such as context and social network data. Such data can be modeled using graphs, and the [...] Read more.
Recommender systems have revolutionized the way users discover and engage with content. Moving beyond the collaborative filtering approach, most modern recommender systems leverage additional sources of information, such as context and social network data. Such data can be modeled using graphs, and the recent advances in Graph Neural Networks have led to the prominence of a new family of graph-based recommender system algorithms. In this work, we propose the RelationalNet algorithm, which not only models user–item, and user–user relationships but also item–item relationships with graphs and uses them as input to the recommendation process. The rationale for utilizing item–item interactions is to enrich the item embeddings by leveraging the similarities between items. By using Graph Neural Networks (GNNs), RelationalNet incorporates social influence and similar item influence into the recommendation process and captures more accurate user interests, especially when traditional methods fall short due to data sparsity. Such models improve the accuracy and effectiveness of recommendation systems by leveraging social connections and item interactions. Results demonstrate that RelationalNet outperforms current state-of-the-art social recommendation algorithms. Full article
(This article belongs to the Special Issue New Trends in Algorithms for Intelligent Recommendation Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>RelationalNet algorithm with <span class="html-italic">k</span> layers. <b>Top part:</b> User–user graph and user–consumed–items graph. <b>Lower part:</b> Item–consumed–users graph and item–item graph.</p> Full article ">Figure 2
<p>Graphs used in the RelationalNet model.</p> Full article ">Figure 3
<p>Embedding layer example for users’ free latent embedding matrix.</p> Full article ">Figure 4
<p>Fusion layer example for a user <span class="html-italic">a</span> is <math display="inline"><semantics> <msubsup> <mi>u</mi> <mi>a</mi> <mn>0</mn> </msubsup> </semantics></math>.</p> Full article ">Figure 5
<p>Diffusion at Node attention layer, i.e., Layer 1 is from 1-hop neighbors, Layer 2 is from 2-hop neighbors, and Layer 3 is from 3-hop neighbors.</p> Full article ">Figure 6
<p>Example of the graph attention at the <math display="inline"><semantics> <msup> <mrow> <mo stretchy="false">(</mo> <mi>k</mi> <mo>+</mo> <mn>1</mn> <mo stretchy="false">)</mo> </mrow> <mrow> <mi>t</mi> <mi>h</mi> </mrow> </msup> </semantics></math> layer for the user embedding <math display="inline"><semantics> <msubsup> <mi>u</mi> <mi>a</mi> <mrow> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> </msubsup> </semantics></math> only.</p> Full article ">Figure 7
<p>Training loss. (<b>a</b>) Training loss of the Diffnet++ model [<a href="#B13-algorithms-16-00515" class="html-bibr">13</a>]. (<b>b</b>) Training loss of the RelationalNet model.</p> Full article ">
28 pages, 1233 KiB  
Article
Trustworthy Digital Representations of Analog Information—An Application-Guided Analysis of a Fundamental Theoretical Problem in Digital Twinning
by Holger Boche, Yannik N. Böck, Ullrich J. Mönich and Frank H. P. Fitzek
Algorithms 2023, 16(11), 514; https://doi.org/10.3390/a16110514 - 9 Nov 2023
Cited by 2 | Viewed by 1521
Abstract
This article compares two methods of algorithmically processing bandlimited time-continuous signals in light of the general problem of finding “suitable” representations of analog information on digital hardware. Albeit abstract, we argue that this problem is fundamental in digital twinning, a signal-processing paradigm the [...] Read more.
This article compares two methods of algorithmically processing bandlimited time-continuous signals in light of the general problem of finding “suitable” representations of analog information on digital hardware. Albeit abstract, we argue that this problem is fundamental in digital twinning, a signal-processing paradigm the upcoming 6G communication-technology standard relies on heavily. Using computable analysis, we formalize a general framework of machine-readable descriptions for representing analytic objects on Turing machines. Subsequently, we apply this framework to sampling and interpolation theory, providing a thoroughly formalized method for digitally processing the information carried by bandlimited analog signals. We investigate discrete-time descriptions, which form the implicit quasi-standard in digital signal processing, and establish continuous-time descriptions that take the signal’s continuous-time behavior into account. Motivated by an exemplary application of digital twinning, we analyze a textbook model of digital communication systems accordingly. We show that technologically fundamental properties, such as a signal’s (Banach-space) norm, can be computed from continuous-time, but not from discrete-time descriptions of the signal. Given the high trustworthiness requirements within 6G, e.g., employed software must satisfy assessment criteria in a provable manner, we conclude that the problem of “trustworthy” digital representations of analog information is indeed essential to near-future information technology. Full article
(This article belongs to the Topic Modeling and Practice for Trustworthy and Secure Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Application of digital twinning as discussed in [<a href="#B10-algorithms-16-00514" class="html-bibr">10</a>]. (<b>Left</b>) A robot (physical entity) is moving on the floor of a laboratory environment. It is sequentially measuring its position relative to a fixed coordinate system and transmitting the relevant data through wireless communication to a receiving end. (<b>Middle</b>) The receiving end tracks the position and forms a virtual representation (digital twin) of the robot inside the room. It updates the virtual representation to match the physical robot whenever new information becomes available. The depicted image is thus a visualization of the robot’s instantaneous machine-readable description. (<b>Right</b>) Using the robot’s virtual representation, the receiving end computes the impulse response of the wireless communication link (the depicted implementation uses a ray-tracing approach), i.e., a sequence of samples representing a bandlimited signal. In essence, the impulse response forms a digital twin of the communication link.</p> Full article ">Figure 2
<p>Generalized Plancherel–Pólya Theorem (c.f. <a href="#sec2-algorithms-16-00514" class="html-sec">Section 2</a>). By evaluating a bandlimited, continuous-time signal (<b>left</b>) at all integer multiples of a suitable sampling interval <span class="html-italic">T</span>, we obtain the sequence of samples (<b>right</b>) corresponding to the function under consideration. We can restore the function through an (infinite) interpolation series, where each term is associated with one of the sample sequence’s components. In particular, each term consists of a time-shifted interpolation function multiplied by the associated sample. This way, the bandlimited function and the sequence of sampling values uniquely determine each other.</p> Full article ">
25 pages, 1338 KiB  
Article
A System to Support Readers in Automatically Acquiring Complete Summarized Information on an Event from Different Sources
by Pietro Dell’Oglio, Alessandro Bondielli and Francesco Marcelloni
Algorithms 2023, 16(11), 513; https://doi.org/10.3390/a16110513 - 8 Nov 2023
Viewed by 1642
Abstract
Today, most newspapers utilize social media to disseminate news. On the one hand, this results in an overload of related articles for social media users. On the other hand, since social media tends to form echo chambers around their users, different opinions and [...] Read more.
Today, most newspapers utilize social media to disseminate news. On the one hand, this results in an overload of related articles for social media users. On the other hand, since social media tends to form echo chambers around their users, different opinions and information may be hidden. Enabling users to access different information (possibly outside of their echo chambers, without the burden of reading entire articles, often containing redundant information) may be a step forward in allowing them to form their own opinions. To address this challenge, we propose a system that integrates Transformer neural models and text summarization models along with decision rules. Given a reference article already read by the user, our system first collects articles related to the same topic from a configurable number of different sources. Then, it identifies and summarizes the information that differs from the reference article and outputs the summary to the user. The core of the system is the sentence classification algorithm, which classifies sentences in the collected articles into three classes based on similarity with the reference article: sentences classified as dissimilar are summarized by using a pre-trained abstractive summarization model. We evaluated the proposed system in two steps. First, we assessed its effectiveness in identifying content differences between the reference article and the related articles by using human judgments obtained through crowdsourcing as ground truth. We obtained an average F1 score of 0.772 against average F1 scores of 0.797 and 0.676 achieved by two state-of-the-art approaches based, respectively, on model tuning and prompt tuning, which require an appropriate tuning phase and, therefore, greater computational effort. Second, we asked a sample of people to evaluate how well the summary generated by the system represents the information that is not present in the article read by the user. The results are extremely encouraging. Finally, we present a use case. Full article
(This article belongs to the Special Issue Machine Learning in Social Network Analytics)
►▼ Show Figures

Figure 1

Figure 1
<p>Flowchart of the proposed system.</p> Full article ">Figure 2
<p>Confusion matrix of our system on the news about Elliot Page coming out.</p> Full article ">Figure 3
<p>Confusion matrix of the fine-tuned BERT on the news about Elliot Page coming out.</p> Full article ">Figure 4
<p>Confusion matrix of the prompt-tuned BERT on the news about Elliot Page coming out.</p> Full article ">Figure 5
<p>Confusion matrix of our system on the news about Italy winning Euro 2020.</p> Full article ">Figure 6
<p>Confusion matrix of the fine-tuned BERT on the news about Italy winning Euro 2020.</p> Full article ">Figure 7
<p>Confusion matrix of the prompt-tuned BERT on the news about Italy winning Euro 2020.</p> Full article ">Figure 8
<p>Average confusion matrix of our system.</p> Full article ">Figure 9
<p>Average confusion matrix of the fine-tuned BERT.</p> Full article ">Figure 10
<p>Average confusion matrix of the prompt-tuned BERT.</p> Full article ">Figure 11
<p>Box plot of human judgments for each news event on the amount of new information contained in the summary, with respect to the reference article.</p> Full article ">Figure 12
<p>Box plot depicting human judgments on the factual correctness of target summaries for each news eventrelative to the reference article.</p> Full article ">Figure 13
<p>Box plot of human judgments for each news event, assessing the information completeness in the summaries of target sentences extracted from the target articles, in comparison to the reference article.</p> Full article ">Figure 14
<p>Interface of the application.</p> Full article ">
3 pages, 193 KiB  
Editorial
Special Issue on Algorithms in Decision Support Systems Vol.2
by Edward Rolando Núñez-Valdez
Algorithms 2023, 16(11), 512; https://doi.org/10.3390/a16110512 - 8 Nov 2023
Viewed by 1210
Abstract
Currently, decision support systems (DSSs) are essential tools that provide information and support for decision making on possible problems that, due to their level of complexity, cannot be easily solved by humans [...] Full article
(This article belongs to the Special Issue Algorithms in Decision Support Systems Vol. 2)
19 pages, 8484 KiB  
Article
A Recommendation System Supporting the Implementation of Sustainable Risk Management Measures in Airport Operations
by Silvia Carpitella, Bruno Brentan, Antonella Certa and Joaquín Izquierdo
Algorithms 2023, 16(11), 511; https://doi.org/10.3390/a16110511 - 7 Nov 2023
Cited by 2 | Viewed by 2264
Abstract
This paper introduces a recommendation system aimed at enhancing the sustainable process of risk management within airport operations, with a special focus on Occupational Stress Risks (OSRs). The recommendation system is implemented via a flexible Python code that offers seamless integration into various [...] Read more.
This paper introduces a recommendation system aimed at enhancing the sustainable process of risk management within airport operations, with a special focus on Occupational Stress Risks (OSRs). The recommendation system is implemented via a flexible Python code that offers seamless integration into various operational contexts. It leverages Fuzzy Cognitive Maps (FCMs) to conduct comprehensive risk assessments, subsequently generating prioritized recommendations for predefined risk management measures aimed at preventing and/or reducing the most critical OSRs. The system’s reliability has been validated by iterating the procedure with diverse input data (i.e., matrices of varying sizes) and measures. This confirms the system’s effectiveness across a broad spectrum of engineering scenarios. Full article
(This article belongs to the Special Issue Mathematical Modelling in Engineering and Human Behaviour)
►▼ Show Figures

Figure 1

Figure 1
<p>Methodological steps.</p> Full article ">Figure 2
<p>Final recommendation.</p> Full article ">Figure 3
<p>Network of relationships.</p> Full article ">Figure 4
<p>Network of relationship–validation [<a href="#B4-algorithms-16-00511" class="html-bibr">4</a>].</p> Full article ">Figure 5
<p>Network of relationship–validation [<a href="#B56-algorithms-16-00511" class="html-bibr">56</a>].</p> Full article ">
21 pages, 1254 KiB  
Article
Detecting and Processing Unsuspected Sensitive Variables for Robust Machine Learning
by Laurent Risser, Agustin Martin Picard, Lucas Hervier and Jean-Michel Loubes
Algorithms 2023, 16(11), 510; https://doi.org/10.3390/a16110510 - 7 Nov 2023
Cited by 1 | Viewed by 1761
Abstract
The problem of algorithmic bias in machine learning has recently gained a lot of attention due to its potentially strong impact on our societies. In much the same manner, algorithmic biases can alter industrial and safety-critical machine learning applications, where high-dimensional inputs are [...] Read more.
The problem of algorithmic bias in machine learning has recently gained a lot of attention due to its potentially strong impact on our societies. In much the same manner, algorithmic biases can alter industrial and safety-critical machine learning applications, where high-dimensional inputs are used. This issue has, however, been mostly left out of the spotlight in the machine learning literature. Contrary to societal applications, where a set of potentially sensitive variables, such as gender or race, can be defined by common sense or by regulations to draw attention to potential risks, the sensitive variables are often unsuspected in industrial and safety-critical applications. In addition, these unsuspected sensitive variables may be indirectly represented as a latent feature of the input data. For instance, the predictions of an image classifier may be altered by reconstruction artefacts in a small subset of the training images. This raises serious and well-founded concerns about the commercial deployment of AI-based solutions, especially in a context where new regulations address bias issues in AI. The purpose of our paper is, then, to first give a large overview of recent advances in robust machine learning. Then, we propose a new procedure to detect and to treat such unknown biases. As far as we know, no equivalent procedure has been proposed in the literature so far. The procedure is also generic enough to be used in a wide variety of industrial contexts. Its relevance is demonstrated on a set of satellite images used to train a classifier. In this illustration, our technique detects that a subset of the training images has reconstruction faults, leading to systematic prediction errors that would have been unsuspected using conventional cross-validation techniques. Full article
(This article belongs to the Topic Modeling and Practice for Trustworthy and Secure Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>General architecture of a neural network designed for classification or regression tasks on images. It first projects the input image information non-linearly into a latent space, and then uses this transformed information for its prediction.</p> Full article ">Figure 2
<p>(<b>Left</b>) Images out of the 10 classes of the EuroSAT dataset. Illustration taken from [<a href="#B5-algorithms-16-00510" class="html-bibr">5</a>]. (<b>Right</b>) Images of the EuroSAT dataset for which the reconstruction in the RGB colour space produced the blue-veil effect.</p> Full article ">Figure 3
<p>Detection of potentially discriminated groups and confirmation of blue-veiled images using a clustering and group-wise performance evaluation methodology. The generalisation properties of a ResNet18 classifier in cluster 2, i.e., for blue-veil images, are particularly lower than in the two other clusters.</p> Full article ">Figure 4
<p>Box-plots of the average errors obtained on the test set with the EuroSat dataset using different models and different training strategies. For each strategy, the two boxplots distinguish the average errors on blue-veiled images (green boxplots) and other images (blue boxplots): (<b>a</b>) Baseline results obtained on different neural network architectures; (<b>b</b>) Effect of different treatments on the average accuracy of the Resnet architecture.</p> Full article ">Figure 5
<p>Detailed convergence of the BCE-loss on all data and on blue-veil images only. Results obtained on the training set (<b>left</b>) and on the test set (<b>right</b>) are represented. Note that the convergence curves obtained on the training set are only represented for the first 40 epochs, and those obtained on the test set are represented on 250 epochs. Five phases <b>A</b> to <b>E</b> are distinguished to discuss the convergence behaviour.</p> Full article ">
19 pages, 4191 KiB  
Article
Parkinson’s Disease Classification Framework Using Vocal Dynamics in Connected Speech
by Sai Bharadwaj Appakaya, Ruchira Pratihar and Ravi Sankar
Algorithms 2023, 16(11), 509; https://doi.org/10.3390/a16110509 - 4 Nov 2023
Cited by 2 | Viewed by 1782
Abstract
Parkinson’s disease (PD) classification through speech has been an advancing field of research because of its ease of acquisition and processing. The minimal infrastructure requirements of the system have also made it suitable for telemonitoring applications. Researchers have studied the effects of PD [...] Read more.
Parkinson’s disease (PD) classification through speech has been an advancing field of research because of its ease of acquisition and processing. The minimal infrastructure requirements of the system have also made it suitable for telemonitoring applications. Researchers have studied the effects of PD on speech from various perspectives using different speech tasks. Typical speech deficits due to PD include voice monotony (e.g., monopitch), breathy or rough quality, and articulatory errors. In connected speech, these symptoms are more emphatic, which is also the basis for speech assessment in popular rating scales used for PD, like the Unified Parkinson’s Disease Rating Scale (UPDRS) and Hoehn and Yahr (HY). The current study introduces an innovative framework that integrates pitch-synchronous segmentation and an optimized set of features to investigate and analyze continuous speech from both PD patients and healthy controls (HC). Comparison of the proposed framework against existing methods has shown its superiority in classification performance and mitigation of overfitting in machine learning models. A set of optimal classifiers with unbiased decision-making was identified after comparing several machine learning models. The outcomes yielded by the classifiers demonstrate that the framework effectively learns the intrinsic characteristics of PD from connected speech, which can potentially offer valuable assistance in clinical diagnosis. Full article
(This article belongs to the Special Issue Machine Learning for Pattern Recognition)
►▼ Show Figures

Figure 1

Figure 1
<p>Methodology Block Diagram.</p> Full article ">Figure 2
<p>Segmentations: Block processing (top) with vertical (red/black) lines showing block limits with block indices (green) and pitch synchronous (bottom). Y–axis shows the normalized signal amplitude.</p> Full article ">Figure 3
<p>Organization of classifiers used in this study.</p> Full article ">Figure 4
<p>Syllabic learning approach showing the original recording (blue) overlaid with voiced sections (red) in the figure at the top. One of the voiced sections, segmented pitch synchronously, is zoomed into and shown at the center.</p> Full article ">Figure 5
<p>Classification results with different segmentations using original labels (MFCCs from Database 1).</p> Full article ">Figure 6
<p>Percentage reduction in classification performance with original and random labels (MFCCs from Database 1).</p> Full article ">Figure 7
<p>Classification performance comparison between MFCCs and PSFs (Database 1).</p> Full article ">Figure 8
<p>Percentage reduction in classification performances with original and random labels—classification between MFCCs and PSF (Database1).</p> Full article ">Figure 9
<p>Classification accuracies with Database 1 used for training and Database 2 used for testing using z-scores for MFCCs and PSF.</p> Full article ">Figure 10
<p>Classification accuracies with Database 1 used for training and Database 2 used for testing using covariances from syllabic analysis z-scores for MFCCs and PSF.</p> Full article ">Figure 11
<p>Performance Metrics with Database 1 for training and Database 2 for testing with (<b>a</b>) MFCCs in three groups; and (<b>b</b>) PSFs in three groups. The three groups are namely group 1 (blue line), group 2 (red line), and group 3 (black line).</p> Full article ">
17 pages, 458 KiB  
Article
Predicting the Gap in the Day-Ahead and Real-Time Market Prices Leveraging Exogenous Weather Data
by Nika Nizharadze, Arash Farokhi Soofi and Saeed Manshadi
Algorithms 2023, 16(11), 508; https://doi.org/10.3390/a16110508 - 4 Nov 2023
Viewed by 1666
Abstract
Predicting the price gap between the day-ahead Market (DAM) and the real-time Market (RTM) plays a vital role in the convergence bidding mechanism of Independent System Operators (ISOs) in wholesale electricity markets. This paper presents a model to predict the values of the [...] Read more.
Predicting the price gap between the day-ahead Market (DAM) and the real-time Market (RTM) plays a vital role in the convergence bidding mechanism of Independent System Operators (ISOs) in wholesale electricity markets. This paper presents a model to predict the values of the price gap between the DAM and RTM using statistical machine learning algorithms and deep neural networks. In this paper, we seek to answer these questions: What will be the impact of predicting the DAM and RTM price gap directly on the prediction performance of learning methods? How can exogenous weather data affect the price gap prediction? In this paper, several exogenous features are collected, and the impacts of these features are examined to capture the best relations between the features and the target variable. An ensemble learning algorithm, namely the Random Forest (RF), is used to select the most important features. A Long Short-Term Memory (LSTM) network is used to capture long-term dependencies in predicting direct gap values between the markets stated. Moreover, the advantages of directly predicting the gap price rather than subtracting the price predictions of the DAM and RTM are shown. The presented results are based on the California Independent System Operator (CAISO)’s electricity market data for two years. The results show that direct gap prediction using exogenous weather features decreases the error of learning methods by 46%. Therefore, the presented method mitigates the prediction error of the price gap between the DAM and RTM. Thus, the convergence bidders can increase their profit, and the ISOs can tune their mechanism accordingly. Full article
(This article belongs to the Special Issue Algorithms for Smart Cities)
►▼ Show Figures

Figure 1

Figure 1
<p>The LSTM architecture of a single unit.</p> Full article ">Figure 2
<p>The probability distribution of the DAM and RTM price predictions for a specific date and time, procured by the RF algorithm.</p> Full article ">Figure 3
<p>A comparison between the probability distribution of direct gap predictions and the difference between separately predicted prices for the DAM and RTM at 8 a.m., procured by LSTM network.</p> Full article ">Figure 4
<p>A comparison between the probability distribution of direct gap predictions and the difference between separately predicted prices for the DAM and RTM at 5 pm, procured by LSTM network.</p> Full article ">Figure 5
<p>A comparison of predicting the gap using LSTM and RF algorithms for the next 96 h to the actual values of the gap.</p> Full article ">Figure 6
<p>A comparison of the relative error of direct gap predictions procured by LSTM and RF algorithms for the next 96 h.</p> Full article ">
20 pages, 2469 KiB  
Article
Deep Dive into Fake News Detection: Feature-Centric Classification with Ensemble and Deep Learning Methods
by Fawaz Khaled Alarfaj and Jawad Abbas Khan
Algorithms 2023, 16(11), 507; https://doi.org/10.3390/a16110507 - 3 Nov 2023
Cited by 4 | Viewed by 2952
Abstract
The online spread of fake news on various platforms has emerged as a significant concern, posing threats to public opinion, political stability, and the dissemination of reliable information. Researchers have turned to advanced technologies, including machine learning (ML) and deep learning (DL) techniques, [...] Read more.
The online spread of fake news on various platforms has emerged as a significant concern, posing threats to public opinion, political stability, and the dissemination of reliable information. Researchers have turned to advanced technologies, including machine learning (ML) and deep learning (DL) techniques, to detect and classify fake news to address this issue. This research study explores fake news classification using diverse ML and DL approaches. We utilized a well-known “Fake News” dataset sourced from Kaggle, encompassing a labelled news collection. We implemented diverse ML models, including multinomial naïve bayes (MNB), gaussian naïve bayes (GNB), Bernoulli naïve Bayes (BNB), logistic regression (LR), and passive aggressive classifier (PAC). Additionally, we explored DL models, such as long short-term memory (LSTM), convolutional neural networks (CNN), and CNN-LSTM. We compared the performance of these models based on key evaluation metrics, such as accuracy, precision, recall, and the F1 score. Additionally, we conducted cross-validation and hyperparameter tuning to ensure optimal performance. The results provide valuable insights into the strengths and weaknesses of each model in classifying fake news. We observed that DL models, particularly LSTM and CNN-LSTM, showed better performance compared to traditional ML models. These models achieved higher accuracy and demonstrated robustness in classification tasks. These findings emphasize the potential of DL models to tackle the spread of fake news effectively and highlight the importance of utilizing advanced techniques to address this challenging problem. Full article
►▼ Show Figures

Figure 1

Figure 1
<p>Proposed methodology for detecting fake news.</p> Full article ">Figure 2
<p>Word cloud of (<b>a</b>) fake news and (<b>b</b>) real news.</p> Full article ">Figure 3
<p>(<b>a</b>–<b>d</b>) top 20 n-grams of the clean and uncleaned corpora.</p> Full article ">Figure 4
<p>Confusion matrix of ensemble and DL models. (<b>A</b>), PA, (<b>B</b>), LR, (<b>C</b>), MNB, (<b>D</b>), BNB, (<b>E</b>), GNN, (<b>F</b>), LSTM, (<b>G</b>), BI-LSTM, (<b>H</b>), CNN, (<b>I</b>) CNN-LSTM, (<b>J</b>) BERT, (<b>K</b>) RoBERTa.</p> Full article ">Figure 4 Cont.
<p>Confusion matrix of ensemble and DL models. (<b>A</b>), PA, (<b>B</b>), LR, (<b>C</b>), MNB, (<b>D</b>), BNB, (<b>E</b>), GNN, (<b>F</b>), LSTM, (<b>G</b>), BI-LSTM, (<b>H</b>), CNN, (<b>I</b>) CNN-LSTM, (<b>J</b>) BERT, (<b>K</b>) RoBERTa.</p> Full article ">Figure 5
<p>ROC curve of ensemble and DL models. (<b>A</b>), PA, (<b>B</b>), LR, (<b>C</b>), MNB, (<b>D</b>), BNB, (<b>E</b>), GNN, (<b>F</b>), LSTM, (<b>G</b>), BI-LSTM, (<b>H</b>), CNN, (<b>I</b>) CNN-LSTM, (<b>J</b>) BERT, (<b>K</b>) RoBERTa.</p> Full article ">
19 pages, 2227 KiB  
Article
Dynamic Demand-Responsive Feeder Bus Network Design for a Short Headway Trunk Line
by Amirreza Nickkar and Young-Jae Lee
Algorithms 2023, 16(11), 506; https://doi.org/10.3390/a16110506 - 31 Oct 2023
Viewed by 1712
Abstract
Recent advancements in technology have increased the potential for demand-responsive feeder transit services to enhance mobility in areas with limited public transit access. For long rail headways, feeder bus network algorithms are straightforward, as the maximum feeder service cycle time is determined by [...] Read more.
Recent advancements in technology have increased the potential for demand-responsive feeder transit services to enhance mobility in areas with limited public transit access. For long rail headways, feeder bus network algorithms are straightforward, as the maximum feeder service cycle time is determined by rail headway, and bus–train matching is unnecessary. However, for short rail headways, the algorithm must address both passenger–feeder-bus and feeder-bus–train matching. This study presents a simulated annealing (SA) algorithm for flexible feeder bus routing, accommodating short headway trunk lines and multiple bus relocations for various stations and trains. A 5 min headway rail trunk line example was utilized to test the algorithm. The algorithm effectively managed bus relocations when optimal routes were infeasible at specific stations. Additionally, the algorithm minimized total costs, accounting for vehicle operating expenses and passenger in-vehicle travel time costs, while considering multiple vehicle relocations. Full article
(This article belongs to the Special Issue Optimization for Vehicle Routing Problems)
►▼ Show Figures

Figure 1

Figure 1
<p>Conceptual operation of the proposed demand-responsive feeder transit service.</p> Full article ">Figure 2
<p>The developed SA algorithm to solve the model.</p> Full article ">Figure 3
<p>The path creator algorithm.</p> Full article ">Figure 4
<p>Geographical distributions of the passengers.</p> Full article ">Figure 5
<p>Illustration for the feeder bus schedules and movements.</p> Full article ">
27 pages, 2365 KiB  
Article
Finding Bottlenecks in Message Passing Interface Programs by Scalable Critical Path Analysis
by Vladimir Korkhov, Ivan Gankevich, Anton Gavrikov, Maria Mingazova, Ivan Petriakov, Dmitrii Tereshchenko, Artem Shatalin and Vitaly Slobodskoy
Algorithms 2023, 16(11), 505; https://doi.org/10.3390/a16110505 - 31 Oct 2023
Viewed by 1631
Abstract
Bottlenecks and imbalance in parallel programs can significantly affect performance of parallel execution. Finding these bottlenecks is a key issue in performance analysis of MPI programs especially on a large scale. One of the ways to discover bottlenecks is to analyze the critical [...] Read more.
Bottlenecks and imbalance in parallel programs can significantly affect performance of parallel execution. Finding these bottlenecks is a key issue in performance analysis of MPI programs especially on a large scale. One of the ways to discover bottlenecks is to analyze the critical path of the parallel program: the longest execution path in the program activity graph. There are a number of methods of finding the critical path; however, most of them suffer a performance drop when scaled. In this paper, we analyze several methods of critical path finding based on classical Dijkstra and Delta-stepping algorithms along with the proposed algorithm based on topological sorting. Corresponding algorithms for each approach are presented including additional enhancements for increasing performance. The implementation of the algorithms and resulting performance for several benchmark applications (NAS Parallel Benchmarks, CP2K, OpenFOAM, LAMMPS, and MiniFE) are analyzed and discussed. Full article
(This article belongs to the Collection Parallel and Distributed Computing: Algorithms and Applications)
►▼ Show Figures

Figure 1

Figure 1
<p>BSP model.</p> Full article ">Figure 2
<p>Part of program activity graph.</p> Full article ">Figure 3
<p>Parallel splits: graph with no splits (<b>left</b>); graph with splits (<b>right</b>).</p> Full article ">Figure 4
<p>Example of topological sorting.</p> Full article ">Figure 5
<p>Groups of overlapping program edges.</p> Full article ">Figure 6
<p>Shuffle.</p> Full article ">Figure 7
<p>Unbalanced (initial) and balanced shuffle data distributions.</p> Full article ">Figure 8
<p>Shuffle time for unbalanced and balanced data distributions.</p> Full article ">Figure 9
<p>Shuffle time for empty and non-empty data requests.</p> Full article ">Figure 10
<p>Results of applying compression: shuffle message size.</p> Full article ">Figure 11
<p>Results of applying compression: shuffle time.</p> Full article ">Figure 12
<p>Topological sorting time dependence on the no. of MPI calls.</p> Full article ">Figure 13
<p>Program activity graphs: program with a collective operation (<b>left</b>); program with sending/receiving to/from neighbour processes using non-blocking point-to-point operations (<b>right</b>).</p> Full article ">Figure 14
<p>NAS Parallel Benchmarks test results (sequential Dijkstra).</p> Full article ">Figure 15
<p>NAS Benchmarks execution time: real time (no critical path data collection).</p> Full article ">Figure 16
<p>NAS Benchmarks execution time: with mpi-graph (parallel Dijkstra algorithm).</p> Full article ">Figure 17
<p>Relative overhead of <span class="html-italic">mpi-graph</span> (parallel Dijkstra).</p> Full article ">Figure 18
<p>OpenFOAM, snappyHexMesh, motorbike test case.</p> Full article ">Figure 19
<p>CP2K, Fayalite test case.</p> Full article ">Figure 20
<p>LAMMPS, lj test case.</p> Full article ">
26 pages, 678 KiB  
Article
Decision-Maker’s Preference-Driven Dynamic Multi-Objective Optimization
by Adekunle Rotimi Adekoya and Mardé Helbig
Algorithms 2023, 16(11), 504; https://doi.org/10.3390/a16110504 - 30 Oct 2023
Cited by 1 | Viewed by 1356
Abstract
Dynamic multi-objective optimization problems (DMOPs) are optimization problems where elements of the problems, such as the objective functions and/or constraints, change with time. These problems are characterized by two or more objective functions, where at least two objective functions are in conflict with [...] Read more.
Dynamic multi-objective optimization problems (DMOPs) are optimization problems where elements of the problems, such as the objective functions and/or constraints, change with time. These problems are characterized by two or more objective functions, where at least two objective functions are in conflict with one another. When solving real-world problems, the incorporation of human decision-makers (DMs)’ preferences or expert knowledge into the optimization process and thereby restricting the search to a specific region of the Pareto-optimal Front (POF) may result in more preferred or suitable solutions. This study proposes approaches that enable DMs to influence the search process with their preferences by reformulating the optimization problems as constrained problems. The subsequent constrained problems are solved using various constraint handling approaches, such as the penalization of infeasible solutions and the restriction of the search to the feasible region of the search space. The proposed constraint handling approaches are compared by incorporating the approaches into a differential evolution (DE) algorithm and measuring the algorithm’s performance using both standard performance measures for dynamic multi-objective optimization (DMOO), as well as newly proposed measures for constrained DMOPs. The new measures indicate how well an algorithm was able to find solutions in the objective space that best reflect the DM’s preferences and the Pareto-optimality goal of dynamic multi-objective optimization algorithms (DMOAs). The results indicate that the constraint handling approaches are effective in finding Pareto-optimal solutions that satisfy the preference constraints of a DM. Full article
(This article belongs to the Special Issue Optimization Algorithms for Decision Support Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Preference-driven search process of a DMOA.</p> Full article ">Figure 2
<p>Decisions found by DPA (above) and RSTFRA (below) for FDA5 with Sphere Spec = <math display="inline"><semantics> <mrow> <mo>(</mo> <mn>1.7808</mn> <mo>,</mo> <mn>2.9185</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>−</mo> <mn>59</mn> </mrow> </msup> <mo>,</mo> <mn>0.0</mn> <mo>,</mo> <mn>1.5</mn> <mo>)</mo> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>n</mi> <mi>t</mi> </msub> <mo>=</mo> <mn>10</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mi>τ</mi> <mi>t</mi> </msub> <mo>=</mo> <mn>5</mn> </mrow> </semantics></math>.</p> Full article ">Figure 3
<p>Decisions found by DPA (above) and PPA (below) for FDA5 with Sphere Spec = <math display="inline"><semantics> <mrow> <mo>(</mo> <mn>1.0276</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>−</mo> <mn>16</mn> </mrow> </msup> <mo>,</mo> <mn>4.4955</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>−</mo> <mn>62</mn> </mrow> </msup> <mo>,</mo> <mn>1.68781</mn> <mo>,</mo> <mn>1.5</mn> <mo>)</mo> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <msub> <mi>n</mi> <mi>t</mi> </msub> <mo>=</mo> <mn>10</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mi>τ</mi> <mi>t</mi> </msub> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>.</p> Full article ">
15 pages, 5285 KiB  
Article
Machine Learning-Based Approach for Predicting Diabetes Employing Socio-Demographic Characteristics
by Md. Ashikur Rahman, Lway Faisal Abdulrazak, Md. Mamun Ali, Imran Mahmud, Kawsar Ahmed and Francis M. Bui
Algorithms 2023, 16(11), 503; https://doi.org/10.3390/a16110503 - 29 Oct 2023
Cited by 1 | Viewed by 2754
Abstract
Diabetes is one of the fatal diseases that play a vital role in the growth of other diseases in the human body. From a clinical perspective, the most significant approach to mitigating the effects of diabetes is early-stage control and management, with the [...] Read more.
Diabetes is one of the fatal diseases that play a vital role in the growth of other diseases in the human body. From a clinical perspective, the most significant approach to mitigating the effects of diabetes is early-stage control and management, with the aim of a potential cure. However, lack of awareness and expensive clinical tests are the primary reasons why clinical diagnosis and preventive measures are neglected in lower-income countries like Bangladesh, Pakistan, and India. From this perspective, this study aims to build an automated machine learning (ML) model, which will predict diabetes at an early stage using socio-demographic characteristics rather than clinical attributes, due to the fact that clinical features are not always accessible to all people from lower-income countries. To find the best fit of the supervised ML classifier of the model, we applied six classification algorithms and found that RF outperformed with an accuracy of 99.36%. In addition, the most significant risk factors were found based on the SHAP value by all the applied classifiers. This study reveals that polyuria, polydipsia, and delayed healing are the most significant risk factors for developing diabetes. The findings indicate that the proposed model is highly capable of predicting diabetes in the early stages. Full article
(This article belongs to the Collection Feature Papers in Algorithms and Mathematical Models for Computer-Assisted Diagnostic Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Experimental methodology of the study for building a diabetes prediction model using socio-demographic characteristics by machine learning techniques.</p> Full article ">Figure 2
<p>Visual representation of K-Fold CV method in supervised machine learning model training and testing.</p> Full article ">Figure 3
<p>Exploratory data analysis result.</p> Full article ">Figure 4
<p>ROC curve and PR curve analysis. (<b>A</b>) ROC curve for the balanced dataset, (<b>B</b>) PR curve for the balanced dataset, (<b>C</b>) ROC curve for the imbalanced dataset, and (<b>D</b>) PR curve for the imbalanced dataset.</p> Full article ">Figure 5
<p>Comparison of the results of different performance metrics based on train–test split and cross-validation for the balanced dataset.</p> Full article ">Figure 6
<p>Accuracy of the six applied classifiers for balanced and imbalanced dataset based on train–test split and cross-validation.</p> Full article ">Figure 7
<p>SHAP summary plot for features’ impact on model prediction. (<b>A</b>) Features’ importance on model prediction by DT, (<b>B</b>) Features’ importance on model prediction by RF, (<b>C</b>) Features’ importance on model prediction by SVM, (<b>D</b>) Features’ importance on model prediction by XGBOOST, (<b>E</b>) Features’ importance on model prediction by LGBM, (<b>F</b>) Features’ importance on model prediction by MLP.</p> Full article ">
26 pages, 5220 KiB  
Article
Automating Stimulation Frequency Selection for SSVEP-Based Brain-Computer Interfaces
by Alexey Kozin, Anton Gerasimov, Maxim Bakaev, Anton Pashkov and Olga Razumnikova
Algorithms 2023, 16(11), 502; https://doi.org/10.3390/a16110502 - 29 Oct 2023
Viewed by 1706
Abstract
Brain–computer interfaces (BCIs) based on steady-state visually evoked potentials (SSVEPs) are inexpensive and do not require user training. However, the highly personalized reaction to visual stimulation is an obstacle to the wider application of this technique, as it can be ineffective, tiring, or [...] Read more.
Brain–computer interfaces (BCIs) based on steady-state visually evoked potentials (SSVEPs) are inexpensive and do not require user training. However, the highly personalized reaction to visual stimulation is an obstacle to the wider application of this technique, as it can be ineffective, tiring, or even harmful at certain frequencies. In our experimental study, we proposed a new approach to the selection of optimal frequencies of photostimulation. By using a custom photostimulation device, we covered a frequency range from 5 to 25 Hz with 1 Hz increments, recording the subjects’ brainwave activity (EEG) and analyzing the signal-to-noise ratio (SNR) changes at the corresponding frequencies. The proposed set of SNR-based coefficients and the discomfort index, determined by the ratio of theta and beta rhythms in the EEG signal, enables the automation of obtaining the recommended stimulation frequencies for use in SSVEP-based BCIs. Full article
(This article belongs to the Special Issue Artificial Intelligence for Medical Imaging)
►▼ Show Figures

Figure 1

Figure 1
<p>Scheme of the EEG cap with Ag/AgCl sintered sensors by MCScap for the 24-channel SmartBCI wireless wearable EEG system.</p> Full article ">Figure 2
<p>The model of the developed photostimulator used for generating photostimuli in this study.</p> Full article ">Figure 3
<p>Reaction speed coefficient values for subject 3.</p> Full article ">Figure 4
<p>Threshold overcoming coefficient values for subject 3.</p> Full article ">Figure 5
<p>Frequency applicability coefficient values for subject 3.</p> Full article ">Figure 6
<p>Mean frequency applicability coefficient values for subject 3.</p> Full article ">Figure 7
<p>Average theta–beta ratio for males and females.</p> Full article ">Figure 8
<p>Discomfort coefficient values for subject 3.</p> Full article ">Figure 9
<p>Main window of the developed application.</p> Full article ">
17 pages, 25202 KiB  
Article
Denoising Diffusion Models on Model-Based Latent Space
by Carmelo Scribano, Danilo Pezzi, Giorgia Franchini and Marco Prato
Algorithms 2023, 16(11), 501; https://doi.org/10.3390/a16110501 - 28 Oct 2023
Viewed by 2360
Abstract
With the recent advancements in the field of diffusion generative models, it has been shown that defining the generative process in the latent space of a powerful pretrained autoencoder can offer substantial advantages. This approach, by abstracting away imperceptible image details and introducing [...] Read more.
With the recent advancements in the field of diffusion generative models, it has been shown that defining the generative process in the latent space of a powerful pretrained autoencoder can offer substantial advantages. This approach, by abstracting away imperceptible image details and introducing substantial spatial compression, renders the learning of the generative process more manageable while significantly reducing computational and memory demands. In this work, we propose to replace autoencoder coding with a model-based coding scheme based on traditional lossy image compression techniques; this choice not only further diminishes computational expenses but also allows us to probe the boundaries of latent-space image generation. Our objectives culminate in the proposal of a valuable approximation for training continuous diffusion models within a discrete space, accompanied by enhancements to the generative model for categorical values. Beyond the good results obtained for the problem at hand, we believe that the proposed work holds promise for enhancing the adaptability of generative diffusion models across diverse data types beyond the realm of imagery. Full article
(This article belongs to the Special Issue Algorithms for Image Processing and Machine Vision)
►▼ Show Figures

Figure 1

Figure 1
<p>Image compression scheme.</p> Full article ">Figure 2
<p>Product quantization. (*) When <math display="inline"><semantics> <mrow> <mi>R</mi> <mo>≠</mo> <mi>I</mi> <mi>d</mi> </mrow> </semantics></math> this corresponds to OPQ.</p> Full article ">Figure 3
<p>This example is obtained with compression schema OPQ-32-8-8, see <a href="#sec4dot1-algorithms-16-00501" class="html-sec">Section 4.1</a> for details. (<b>a</b>) <math display="inline"><semantics> <mrow> <mi mathvariant="script">D</mi> <mo>(</mo> <mi mathvariant="script">E</mi> <mo>(</mo> <mi>x</mi> <mo>)</mo> <mo>)</mo> </mrow> </semantics></math>—only reconstruction noise. (<b>b</b>) <math display="inline"><semantics> <mrow> <mi mathvariant="script">D</mi> <mo>(</mo> <mi mathvariant="script">E</mi> <mo>(</mo> <mi>x</mi> <mo>)</mo> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </semantics></math> before rearranging <math display="inline"><semantics> <mi mathvariant="script">I</mi> </semantics></math>. (<b>c</b>) <math display="inline"><semantics> <mrow> <mi mathvariant="script">D</mi> <mo>(</mo> <mi mathvariant="script">E</mi> <mo>(</mo> <mi>x</mi> <mo>)</mo> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </semantics></math> after rearranging <math display="inline"><semantics> <mi mathvariant="script">I</mi> </semantics></math>.</p> Full article ">Figure 4
<p>Effect of temperature smoothing of categorical distribution.</p> Full article ">Figure 5
<p>(<b>a</b>) Generated Image. (<b>b</b>) After Refinement.</p> Full article ">Figure 6
<p>Samples generated with continuous formulation (after refiner).</p> Full article ">Figure 7
<p>Sampling steps (<math display="inline"><semantics> <mrow> <mi>t</mi> <mo>=</mo> <mn>1000</mn> <mo>,</mo> <mn>900</mn> <mo>,</mo> <mo>…</mo> <mo>.</mo> <mo>,</mo> <mn>0</mn> </mrow> </semantics></math>) of the model trained with the categorical parameterization of the reverse process.</p> Full article ">Figure 8
<p>Samples generated with categorical formulation (no refiner needed).</p> Full article ">Figure 9
<p>Effects of distribution smoothing on <math display="inline"><semantics> <msub> <mover accent="true"> <mi>z</mi> <mo stretchy="false">^</mo> </mover> <mn>0</mn> </msub> </semantics></math>. (<b>a</b>) argmax. (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>γ</mi> <mo>=</mo> <mn>0.1</mn> </mrow> </semantics></math> (<b>c</b>) <math display="inline"><semantics> <mrow> <mi>γ</mi> <mo>=</mo> <mn>1.0</mn> </mrow> </semantics></math>. (<b>d</b>) <math display="inline"><semantics> <mrow> <mi>γ</mi> <mo>=</mo> <mn>2.0</mn> </mrow> </semantics></math>.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-36
Previous Issue
Next Issue
Back to TopTop