Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.50
Mahdi Asadi, Ali x Ali Rezaei, A. Abazari
Wireless power transmission systems enable the transfer of electricity between grids without the use of physical wires. Different methods are employed for wireless power transfer, each suited to different distances. Inductive coupling, the subject of this study, is typically used for shorter distances. The effectiveness of inductive coupling systems is evaluated using a parameter called mutual inductance. In the present study, an attempt is made to provide a model for calculating mutual inductance in wireless power transfer systems using a machine learning approach. To achieve this goal, finite element simulations are employed, and 64 datasets are generated from mutual inductance calculations in various scenarios. These datasets are used to train machine learning regression algorithms, including linear regression, support vector regression, decision tree regression, and artificial neural networks. The evaluation results, using performance metrics such as R-squared, mean absolute error, and root mean square error, confirm that among these four algorithms, the artificial neural network exhibits higher computational accuracy with an R-squared value of 0.950 for predicting test data.
无线输电系统使电网之间的电力传输无需使用物理电线。无线电力传输采用不同的方法，每种方法适用于不同的距离。电感耦合是本研究的主题，通常用于较短的距离。电感耦合系统的有效性是通过一个称为互感的参数来评估的。在本研究中，我们尝试使用机器学习方法提供一个模型，用于计算无线电力传输系统中的互感。为实现这一目标，我们采用了有限元模拟，并从各种情况下的互感计算中生成了 64 个数据集。这些数据集用于训练机器学习回归算法，包括线性回归、支持向量回归、决策树回归和人工神经网络。使用 R 平方、平均绝对误差和均方根误差等性能指标进行的评估结果证实，在这四种算法中，人工神经网络的计算精度更高，预测测试数据的 R 平方值为 0.950。
{"title":"Calculation of Mutual Inductance between Two Planar Coils with Custom Specifications and Positions Using a Machine Learning Approach","authors":"Mahdi Asadi, Ali x Ali Rezaei, A. Abazari","doi":"10.58190/icontas.2023.50","DOIUrl":"https://doi.org/10.58190/icontas.2023.50","url":null,"abstract":"Wireless power transmission systems enable the transfer of electricity between grids without the use of physical wires. Different methods are employed for wireless power transfer, each suited to different distances. Inductive coupling, the subject of this study, is typically used for shorter distances. The effectiveness of inductive coupling systems is evaluated using a parameter called mutual inductance. In the present study, an attempt is made to provide a model for calculating mutual inductance in wireless power transfer systems using a machine learning approach. To achieve this goal, finite element simulations are employed, and 64 datasets are generated from mutual inductance calculations in various scenarios. These datasets are used to train machine learning regression algorithms, including linear regression, support vector regression, decision tree regression, and artificial neural networks. The evaluation results, using performance metrics such as R-squared, mean absolute error, and root mean square error, confirm that among these four algorithms, the artificial neural network exhibits higher computational accuracy with an R-squared value of 0.950 for predicting test data.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139193856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.58
Naokix NaokiKANEKO, Takuya Nakagawa, Daiki Asada, H. Oya
This paper shows a new parameter adjustment law of PID controllers for MIMO linear system with guaranteed disturbance attenuation performance. In the proposed design approach, the design problem of PID parameters is reduced to the problem of static output feedback controllers for MIMO linear systems. In this paper, we show that sufficient conditions for the existence of the proposed PID control system can be reduced to solvability of linear matrix inequalities (LMIs). Finally, a simple numerical example is shown to effectiveness of the proposed PID control system.
本文展示了一种新的多输入多输出线性系统 PID 控制器参数调整法，其扰动衰减性能得到了保证。在所提出的设计方法中，PID 参数的设计问题被简化为 MIMO 线性系统的静态输出反馈控制器问题。本文表明，拟议 PID 控制系统存在的充分条件可以简化为线性矩阵不等式（LMI）的可解性。最后，通过一个简单的数值示例说明了所提出的 PID 控制系统的有效性。
{"title":"A PARAMETER ADJUSTMENT LAW FOR MULTIVARIABLE PID CONTROL SYSTEMS WITH DISTURBANCE ATTENUATION PERFORMANCE","authors":"Naokix NaokiKANEKO, Takuya Nakagawa, Daiki Asada, H. Oya","doi":"10.58190/icontas.2023.58","DOIUrl":"https://doi.org/10.58190/icontas.2023.58","url":null,"abstract":"This paper shows a new parameter adjustment law of PID controllers for MIMO linear system with guaranteed disturbance attenuation performance. In the proposed design approach, the design problem of PID parameters is reduced to the problem of static output feedback controllers for MIMO linear systems. In this paper, we show that sufficient conditions for the existence of the proposed PID control system can be reduced to solvability of linear matrix inequalities (LMIs). Finally, a simple numerical example is shown to effectiveness of the proposed PID control system.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139194534","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.56
Tatsuya Komatsu, Hiroto Noma, Takumi Asaoka, H. Oya, R. Miura, Koji Yoshioka
In this paper, we show an efficient feature analysis method of body surface temperature (ST) data so as to develop accurate prediction systems for artificial insemination (AI) timing of cattle. In the proposed analysis method, by using the fundamental waveform synthesis method based on the Fourier transform, approximate waveforms for the target waveform were derived. Additionally, reconstructed waveforms which does not correspond to both high frequency noise and circadian rhythm were generated. The two reconstructed waveforms derived from the approximate waveforms were used to predict the optimal AI timing and to discriminate the normal phase, respectively.
{"title":"AN EFFICIENT FEATURE ANALYSIS METHOD OF BIOLOGICAL DATA FOR IMPROVING CATTLE CONCEPTION RATE","authors":"Tatsuya Komatsu, Hiroto Noma, Takumi Asaoka, H. Oya, R. Miura, Koji Yoshioka","doi":"10.58190/icontas.2023.56","DOIUrl":"https://doi.org/10.58190/icontas.2023.56","url":null,"abstract":"In this paper, we show an efficient feature analysis method of body surface temperature (ST) data so as to develop accurate prediction systems for artificial insemination (AI) timing of cattle. In the proposed analysis method, by using the fundamental waveform synthesis method based on the Fourier transform, approximate waveforms for the target waveform were derived. Additionally, reconstructed waveforms which does not correspond to both high frequency noise and circadian rhythm were generated. The two reconstructed waveforms derived from the approximate waveforms were used to predict the optimal AI timing and to discriminate the normal phase, respectively.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139193401","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We have developed a methodology that seeks to associate the molecularity of compounds with the perceptions of specific odor or taste. This methodology goes beyond gross structural features for a molecule: aromatic or aliphatic rings, lengths of the aliphatic straight chains, or the nature and variation in the functional groups. We target specific atom pairs–bonded or remote–within the smell and taste molecule that have structural-electronic features that are reproducible across molecules that elicit similar smell and taste responses. We represent the “structure” of the atom pair by its interatomic distance. The “electronic” aspects are represented by Nuclear Magnetic Resonance (NMR) chemical shifts that uniquely define the electronic environments of the atoms. We used quantum chemistry calculations and the density functional theory (DFT) to determine the chemical shifts and interatomic distances (through the Z-matrix). We used this methodology to process 19 molecules that elicited the smell of “brown,” and 18 molecules that elicited the taste of “brown.” These molecules were accessed through odor and taste indices from the GoodScentsCompany resource (https://www.thegoodscentscompany.com/). These “brown” odorants and tastants elicited other associated smells and tastes. We identified and illustrated specific bond pairs that elicited different smells and tastes. While smell and taste are intrinsically related, our studies also show atom pairs that are likely responsible exclusively for smell and taste, as well as pairs that elicit both. This work will be impactful in the domain of drug design in the pharmaceutical industry, in addition to enhancing our understanding of how a chemical catalyzes the process that results in chemosensory perception.
{"title":"Exploring the Molecularity of the Odor and Taste Perceptions of “Brown”: A Computational Approach","authors":"Hirva Bhayani, Roshan Thilakarathne, Neranjan Perera, Chiquito Crasto","doi":"10.58190/icontas.2023.61","DOIUrl":"https://doi.org/10.58190/icontas.2023.61","url":null,"abstract":"We have developed a methodology that seeks to associate the molecularity of compounds with the perceptions of specific odor or taste. This methodology goes beyond gross structural features for a molecule: aromatic or aliphatic rings, lengths of the aliphatic straight chains, or the nature and variation in the functional groups. We target specific atom pairs–bonded or remote–within the smell and taste molecule that have structural-electronic features that are reproducible across molecules that elicit similar smell and taste responses. We represent the “structure” of the atom pair by its interatomic distance. The “electronic” aspects are represented by Nuclear Magnetic Resonance (NMR) chemical shifts that uniquely define the electronic environments of the atoms. We used quantum chemistry calculations and the density functional theory (DFT) to determine the chemical shifts and interatomic distances (through the Z-matrix). We used this methodology to process 19 molecules that elicited the smell of “brown,” and 18 molecules that elicited the taste of “brown.” These molecules were accessed through odor and taste indices from the GoodScentsCompany resource (https://www.thegoodscentscompany.com/). These “brown” odorants and tastants elicited other associated smells and tastes. We identified and illustrated specific bond pairs that elicited different smells and tastes. While smell and taste are intrinsically related, our studies also show atom pairs that are likely responsible exclusively for smell and taste, as well as pairs that elicit both. This work will be impactful in the domain of drug design in the pharmaceutical industry, in addition to enhancing our understanding of how a chemical catalyzes the process that results in chemosensory perception.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139194308","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.55
Yuta Yoshikawa, Takayuki Okai, H. Oya, Minoru Yoshida, Md.Masudur Rahman
In this paper, we propose a recognition method of R-peaks on electrocardiograms (ECGs) based on wavelet transform with pseudo-differential operators. It is well known that the accurate recognition of R-peaks is highly importance for diagnosis of cardiac diseases and autonomic ataxia. However, the existing results for detection of R-peaks are not always accurate and can have missed peaks or false. Difficulties in accurate R-peaks detection is caused by presence of various noises in ECGs and the physiological variability of the QRS complex. From the above, we propose a more flexible and adaptive recognition method of R-peaks. In order to develop the proposed detection method, noises, artifacts, and baseline variation in ECGs are firstly suppressed by using the low-pass/high-pass filters, moving average, and MaMeMi filter. Next, the time-frequency domain's energy distribution is computed by using wavelet transform with pseudo-differential operators. Furthermore, we introduce a time-series index, -Normalized Spectrum Index ( f^p-NSI) obtained by scalograms based on the wavelet transform with pseudo-differential operators. Finally, R-peaks are recognized by taking the threshold toward the results of f^p-NSI. In this paper, we present the proposed recognition method of R-peaks on ECGs, and the effectiveness (accuracy) of the proposed method is evaluated.
本文提出了一种基于小波变换和伪微分算子的心电图（ECG）R 峰识别方法。众所周知，准确识别 R 峰对于诊断心脏疾病和自主神经共济失调非常重要。然而，现有的 R 峰检测结果并不总是准确的，可能会出现漏峰或假峰。心电图中存在的各种噪声和 QRS 波群的生理变化是造成 R 峰难以准确检测的原因。综上所述，我们提出了一种更加灵活和自适应的 R 峰识别方法。为了开发所提出的检测方法，首先使用低通/高通滤波器、移动平均滤波器和 MaMeMi 滤波器抑制心电图中的噪声、伪像和基线变化。然后，使用带伪差分算子的小波变换计算时频域的能量分布。此外，我们还引入了一种时间序列指数--归一化频谱指数（f^p-NSI），该指数由基于伪微分算子的小波变换得到。最后，通过对 f^p-NSI 的结果取阈值来识别 R 峰。本文提出了在心电图上识别 R 峰的方法，并对该方法的有效性（准确性）进行了评估。
{"title":"A RECOGNITION METHOD OF R-PEAKS ON ELECTROCARDIOGRAMS BASED ON WAVELET TRANSFORM WITH PSEUDO-DIFFERENTIAL OPERATORS","authors":"Yuta Yoshikawa, Takayuki Okai, H. Oya, Minoru Yoshida, Md.Masudur Rahman","doi":"10.58190/icontas.2023.55","DOIUrl":"https://doi.org/10.58190/icontas.2023.55","url":null,"abstract":"In this paper, we propose a recognition method of R-peaks on electrocardiograms (ECGs) based on wavelet transform with pseudo-differential operators. It is well known that the accurate recognition of R-peaks is highly importance for diagnosis of cardiac diseases and autonomic ataxia. However, the existing results for detection of R-peaks are not always accurate and can have missed peaks or false. Difficulties in accurate R-peaks detection is caused by presence of various noises in ECGs and the physiological variability of the QRS complex. From the above, we propose a more flexible and adaptive recognition method of R-peaks. In order to develop the proposed detection method, noises, artifacts, and baseline variation in ECGs are firstly suppressed by using the low-pass/high-pass filters, moving average, and MaMeMi filter. Next, the time-frequency domain's energy distribution is computed by using wavelet transform with pseudo-differential operators. Furthermore, we introduce a time-series index, -Normalized Spectrum Index ( f^p-NSI) obtained by scalograms based on the wavelet transform with pseudo-differential operators. Finally, R-peaks are recognized by taking the threshold toward the results of f^p-NSI. In this paper, we present the proposed recognition method of R-peaks on ECGs, and the effectiveness (accuracy) of the proposed method is evaluated.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139195650","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.57
K. Kapasakalis, Spiridon Kapasakalis, Evangelos x Evangelos SAPOUNTZAKIS
In this research study, the KDamper concept is extended (EKD device) and applied to multiple floors of existing multi-story building structures, aiming to reduce the structure dynamic responses due to earthquake excitations. The KDamper is a novel passive vibration absorption concept, based essentially on the optimal combination of appropriate stiffness elements, one of which has a negative value (NS). The mass requirements of KDamper are reduced, compared to the Tuned Mass Damper (TMD), as the NS element is implemented to the installed mass and the NS force is in phase with the inertia force, artificially amplifying it. Inspired by the concept of distributed TMDs (d-TMDs), multiple EKDs (d-EKDs) are installed and distributed along the height of the structure, for seismic protection. The design and spatial allocation of these EKDs are determined using a Harmony Search (HS) algorithm, which identifies optimal device parameters while adhering to structural constraints and limitations. Artificial accelerograms are generated and introduced as input to the optimization process. Based on the numerical results obtained, the d-EKD concept, outperforms the d-TMD in reducing the structural dynamic responses, introducing one order of magnitude smaller added oscillating masses. In addition, results indicate no significant alteration of the structural properties and eigenfrequencies due to the installation of the proposed EKD devices, despite the addition of masses and NS elements.
{"title":"SEISMIC PROTECTION OF EXISTING STRUCTURES WITH DISTRIBUTED NEGATIVE STIFFNESS DEVICES","authors":"K. Kapasakalis, Spiridon Kapasakalis, Evangelos x Evangelos SAPOUNTZAKIS","doi":"10.58190/icontas.2023.57","DOIUrl":"https://doi.org/10.58190/icontas.2023.57","url":null,"abstract":"In this research study, the KDamper concept is extended (EKD device) and applied to multiple floors of existing multi-story building structures, aiming to reduce the structure dynamic responses due to earthquake excitations. The KDamper is a novel passive vibration absorption concept, based essentially on the optimal combination of appropriate stiffness elements, one of which has a negative value (NS). The mass requirements of KDamper are reduced, compared to the Tuned Mass Damper (TMD), as the NS element is implemented to the installed mass and the NS force is in phase with the inertia force, artificially amplifying it. Inspired by the concept of distributed TMDs (d-TMDs), multiple EKDs (d-EKDs) are installed and distributed along the height of the structure, for seismic protection. The design and spatial allocation of these EKDs are determined using a Harmony Search (HS) algorithm, which identifies optimal device parameters while adhering to structural constraints and limitations. Artificial accelerograms are generated and introduced as input to the optimization process. Based on the numerical results obtained, the d-EKD concept, outperforms the d-TMD in reducing the structural dynamic responses, introducing one order of magnitude smaller added oscillating masses. In addition, results indicate no significant alteration of the structural properties and eigenfrequencies due to the installation of the proposed EKD devices, despite the addition of masses and NS elements.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139190278","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.60
Takumi Asaoka, Hiroto Noma, Tatsuya Komatsu, H. Oya, R. Miura, Koji Yoshioka
For farmers who maintain dairy cattle, artificial insemination (AI) is one of important events in cattle, because it may lead to lose money by missing out on AI. However, the accuracy for detection depends on the time and number of observations when the estrus behavior and signs for cattle during the estrus season are visually assessed by experts and farmers, and the detection accuracy via experts and farmers is approximately 60%. For farmers, it is obvious that improving reproductive efficiency can save time and money. Therefore, various detection strategies for AI timing such as pedometers and methods based on observation of hormone in estrus have been well studied. Additionally, a detection strategy based on variations for temperature corresponding to ovulation has also been presented. In particular, the accuracy of detection of AI timing based on monitoring the vaginal temperature is greater than that for other methods such as pedometer and so on, i.e., it seems that an optimal timing of AI based on vaginal temperature in cattle is more effective. Although there are some existing results for detection of AI timing based on vaginal temperature and vaginal electrical resistance data, further improvement of accuracy is required in practical use. In this paper, we propose an estimation method for the optimal AI timing by analyzing both vaginal temperature and vaginal electrical resistance data. In our approach, as preprocessing, MaMeMi filter and Gaussian kernel smoother are newly introduced for the purpose of reducing the effect of circadian rhythms and various noises. Moreover, we adopt continuous wavelet transformation to analyze biological data, and NSI (Normalized Spectrum Index) is calculated. Finally, the optimal timing for AI can be estimated by using the Mahalanobis distance. In this paper, we present the proposed estimation algorithm and evaluate the proposed approach.
{"title":"ANALYSIS OF BIOLOGICAL DATA OF CATTLE AND WAVELET TRANSFORM BASED PREDICTION FOR OPTIMAL INSEMINATION PHASE","authors":"Takumi Asaoka, Hiroto Noma, Tatsuya Komatsu, H. Oya, R. Miura, Koji Yoshioka","doi":"10.58190/icontas.2023.60","DOIUrl":"https://doi.org/10.58190/icontas.2023.60","url":null,"abstract":"For farmers who maintain dairy cattle, artificial insemination (AI) is one of important events in cattle, because it may lead to lose money by missing out on AI. However, the accuracy for detection depends on the time and number of observations when the estrus behavior and signs for cattle during the estrus season are visually assessed by experts and farmers, and the detection accuracy via experts and farmers is approximately 60%. For farmers, it is obvious that improving reproductive efficiency can save time and money. Therefore, various detection strategies for AI timing such as pedometers and methods based on observation of hormone in estrus have been well studied. Additionally, a detection strategy based on variations for temperature corresponding to ovulation has also been presented. In particular, the accuracy of detection of AI timing based on monitoring the vaginal temperature is greater than that for other methods such as pedometer and so on, i.e., it seems that an optimal timing of AI based on vaginal temperature in cattle is more effective. Although there are some existing results for detection of AI timing based on vaginal temperature and vaginal electrical resistance data, further improvement of accuracy is required in practical use. In this paper, we propose an estimation method for the optimal AI timing by analyzing both vaginal temperature and vaginal electrical resistance data. In our approach, as preprocessing, MaMeMi filter and Gaussian kernel smoother are newly introduced for the purpose of reducing the effect of circadian rhythms and various noises. Moreover, we adopt continuous wavelet transformation to analyze biological data, and NSI (Normalized Spectrum Index) is calculated. Finally, the optimal timing for AI can be estimated by using the Mahalanobis distance. In this paper, we present the proposed estimation algorithm and evaluate the proposed approach.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139193816","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.51
Muhammet Fatih Aslan
Skin diseases are one of the most common diseases in humans. Due to its many various symptoms and types, computer vision studies have been frequently applied to its diagnosis and classification. Previous studies have frequently used machine learning methods and deep learning-based Convolutional Neural Networks (CNN) for skin disease diagnosis. Although deep learning-based applications have achieved great success in terms of detection accuracy, research continues to ensure the desired performance. However, Vision Transformer (ViT), recently proposed as a competitive alternative to CNNs, is gaining increasing popularity. This paper compares ResNet18 and ResNet50 networks, important CNN models, with ViT for classifying skin disease. The comparison is applied on a dataset containing a small number of samples. In the application performed on a dataset containing skin disease images, ViT provides 68.93% classification accuracy, while ResNet18 and ResNet50 classification accuracy is 61.65% and 61.17%, respectively. Other metrics calculated along with the accuracies also prove the superiority of ViT over ResNet models. However, ViT has a big disadvantage in terms of training time.
{"title":"COMPARISON OF VISION TRANSFORMERS AND CONVOLUTIONAL NEURAL NETWORKS FOR SKIN DISEASE CLASSIFICATION","authors":"Muhammet Fatih Aslan","doi":"10.58190/icontas.2023.51","DOIUrl":"https://doi.org/10.58190/icontas.2023.51","url":null,"abstract":"Skin diseases are one of the most common diseases in humans. Due to its many various symptoms and types, computer vision studies have been frequently applied to its diagnosis and classification. Previous studies have frequently used machine learning methods and deep learning-based Convolutional Neural Networks (CNN) for skin disease diagnosis. Although deep learning-based applications have achieved great success in terms of detection accuracy, research continues to ensure the desired performance. However, Vision Transformer (ViT), recently proposed as a competitive alternative to CNNs, is gaining increasing popularity. This paper compares ResNet18 and ResNet50 networks, important CNN models, with ViT for classifying skin disease. The comparison is applied on a dataset containing a small number of samples. In the application performed on a dataset containing skin disease images, ViT provides 68.93% classification accuracy, while ResNet18 and ResNet50 classification accuracy is 61.65% and 61.17%, respectively. Other metrics calculated along with the accuracies also prove the superiority of ViT over ResNet models. However, ViT has a big disadvantage in terms of training time.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139190726","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.54
E. M. Abdelkader, N. Elshaboury, Eslam Ali, Ghasan Alfalah, Ahmed Mansour, Abobakr Alsakkaf
The huge increase in energy consumption in recent decades, has made it cumbersome to anticipate energy usage in the residential sector. However, despite substantial advancements in computation and simulation, the modelling of residential building energy use is still in need of improvement for efficient and reliable solutions. To this end, the overarching objective of this research study is to construct a self-adaptive model (HBO-DL) for predicting the amounts of heating and cooling loads in residential buildings. The developed HBO-DL model is envisioned on coupling Bayesian optimization with deep learning neural network. Five statistical metrics of mean absolute percentage error (MAPE), root mean squared error (RMSE), root mean squared logarithmic error (RMSLE), mean absolute error (MAE) and normalized root mean squared error (NRMSE), are leveraged to measure and test the accuracies of the developed HBO-DL. Analytical results explicated that the developed HBO-DL model can endorse informed decision-making and foster energy conservation in built environment.
{"title":"A Hyper Parametrized Deep Learning Model for Analyzing Heating and Cooling Loads in Energy Efficient Buildings","authors":"E. M. Abdelkader, N. Elshaboury, Eslam Ali, Ghasan Alfalah, Ahmed Mansour, Abobakr Alsakkaf","doi":"10.58190/icontas.2023.54","DOIUrl":"https://doi.org/10.58190/icontas.2023.54","url":null,"abstract":"The huge increase in energy consumption in recent decades, has made it cumbersome to anticipate energy usage in the residential sector. However, despite substantial advancements in computation and simulation, the modelling of residential building energy use is still in need of improvement for efficient and reliable solutions. To this end, the overarching objective of this research study is to construct a self-adaptive model (HBO-DL) for predicting the amounts of heating and cooling loads in residential buildings. The developed HBO-DL model is envisioned on coupling Bayesian optimization with deep learning neural network. Five statistical metrics of mean absolute percentage error (MAPE), root mean squared error (RMSE), root mean squared logarithmic error (RMSLE), mean absolute error (MAE) and normalized root mean squared error (NRMSE), are leveraged to measure and test the accuracies of the developed HBO-DL. Analytical results explicated that the developed HBO-DL model can endorse informed decision-making and foster energy conservation in built environment.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139187765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-01DOI: 10.58190/icontas.2023.48
Elham Tahsin Yasin, Murat Koklu
An analysis of the "Pudina Leaf Dataset: Freshness Analysis" reveals distinct classes of dried, fresh, and spoiled mint leaves. Convolutional neural networks, InceptionV3 and VGG19, were used to extract features from the dataset using advanced image processing techniques. The classification task was then performed using a Random Forest machine learning algorithm. In this study, notable results were obtained, proving the effectiveness of the selected methodologies. Mint (Pudina) leaves were classified accurately using InceptionV3-extracted features at 94.8%, demonstrating robust performance in distinguishing freshness states. This deep learning architecture was further shown to be able to capture meaningful patterns within the dataset by utilizing VGG19-extracted features, resulting in an improved accuracy of 96.8%.
{"title":"Utilizing Random Forests for the Classification of Pudina Leaves through Feature Extraction with InceptionV3 and VGG19","authors":"Elham Tahsin Yasin, Murat Koklu","doi":"10.58190/icontas.2023.48","DOIUrl":"https://doi.org/10.58190/icontas.2023.48","url":null,"abstract":"An analysis of the \"Pudina Leaf Dataset: Freshness Analysis\" reveals distinct classes of dried, fresh, and spoiled mint leaves. Convolutional neural networks, InceptionV3 and VGG19, were used to extract features from the dataset using advanced image processing techniques. The classification task was then performed using a Random Forest machine learning algorithm. In this study, notable results were obtained, proving the effectiveness of the selected methodologies. Mint (Pudina) leaves were classified accurately using InceptionV3-extracted features at 94.8%, demonstrating robust performance in distinguishing freshness states. This deep learning architecture was further shown to be able to capture meaningful patterns within the dataset by utilizing VGG19-extracted features, resulting in an improved accuracy of 96.8%.","PeriodicalId":509439,"journal":{"name":"Proceedings of the International Conference on New Trends in Applied Sciences","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139196209","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：