Pub Date : 2022-12-20DOI: 10.13164/mendel.2022.2.054
Werry Febrianti, K. A. Sidarto, N. Sumarti
This paper examines the implementation of simple combination mutation of differential evolution algorithm for solving stiff ordinary differential equations. We use the weighted residual method with a series expansion to approximate the solutions of stiff ordinary differential equations. We solve the problems from an ordinary stiff differential equation for linear and nonlinear problems. Then, we also implement our method for solving stiff systems of ordinary differential equations. We find that our algorithm can approximate the exact solution of a stiff ordinary differential equation with the smallest error for each length of series that we have chosen. Thus, this approximation method, by using the optimization method of simple combination differential evolution, can be a good tool for solving stiff ordinary differential equations.
{"title":"An Approximate Optimization Method for Solving Stiff Ordinary Differential Equations With Combinational Mutation Strategy of Differential Evolution Algorithm","authors":"Werry Febrianti, K. A. Sidarto, N. Sumarti","doi":"10.13164/mendel.2022.2.054","DOIUrl":"https://doi.org/10.13164/mendel.2022.2.054","url":null,"abstract":"This paper examines the implementation of simple combination mutation of differential evolution algorithm for solving stiff ordinary differential equations. We use the weighted residual method with a series expansion to approximate the solutions of stiff ordinary differential equations. We solve the problems from an ordinary stiff differential equation for linear and nonlinear problems. Then, we also implement our method for solving stiff systems of ordinary differential equations. We find that our algorithm can approximate the exact solution of a stiff ordinary differential equation with the smallest error for each length of series that we have chosen. Thus, this approximation method, by using the optimization method of simple combination differential evolution, can be a good tool for solving stiff ordinary differential equations.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"23 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73321574","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 : 2022-12-20DOI: 10.13164/mendel.2022.2.041
H. T. Minh, T. P. Anh, Van Nguyen Nhan
One of the essential aspects of smart farming and precision agriculture is quickly and accurately identifying diseases. Utilizing plant imaging and recently developed machine learning algorithms, the timely detection of diseases provides many benefits to farmers regarding crop and product quality. Specifically, for farmers in remote areas, disease diagnostics on edge devices is the most effective and optimal method to handle crop damage as quickly as possible. However, the limitations posed by the equipment’s limited resources have reduced the accuracy of disease detection. Consequently, adopting an efficient machine-learning model and decreasing the model size to fit the edge device is an exciting problem that receives significant attention from researchers and developers. This work takes advantage of previous research on deep learning model performance evaluation to present a model that applies to both the Plant-Village laboratory dataset and the Plant-Doc natural-type dataset. The evaluation results indicate that the proposed model is as effective as the current state-of-the-art model. Moreover, due to the quantization technique, the system performance stays the same when the model size is reduced to accommodate the edge device.
{"title":"A Novel Light-Weight DCNN Model for Classifying Plant Diseases on Internet of Things Edge Devices","authors":"H. T. Minh, T. P. Anh, Van Nguyen Nhan","doi":"10.13164/mendel.2022.2.041","DOIUrl":"https://doi.org/10.13164/mendel.2022.2.041","url":null,"abstract":"One of the essential aspects of smart farming and precision agriculture is quickly and accurately identifying diseases. Utilizing plant imaging and recently developed machine learning algorithms, the timely detection of diseases provides many benefits to farmers regarding crop and product quality. Specifically, for farmers in remote areas, disease diagnostics on edge devices is the most effective and optimal method to handle crop damage as quickly as possible. However, the limitations posed by the equipment’s limited resources have reduced the accuracy of disease detection. Consequently, adopting an efficient machine-learning model and decreasing the model size to fit the edge device is an exciting problem that receives significant attention from researchers and developers. This work takes advantage of previous research on deep learning model performance evaluation to present a model that applies to both the Plant-Village laboratory dataset and the Plant-Doc natural-type dataset. The evaluation results indicate that the proposed model is as effective as the current state-of-the-art model. Moreover, due to the quantization technique, the system performance stays the same when the model size is reduced to accommodate the edge device.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"52 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91257539","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 : 2022-12-20DOI: 10.13164/mendel.2022.2.097
Petr Soustek, R. Matousek, J. Dvorak, Lenka Manakova
Path planning or network route planning problems are an important issue in AI, robotics, or computer games. Appropriate implementation and knowledge of advanced and classical path-planning algorithms can be important for both autonomous navigation systems and computer games. In this paper, we compare advanced path planning algorithms implemented on a two-dimensional grid. Advanced path planning algorithms, including pseudocode, are introduced. The experiments were performed in the Python environment, thus with a significant performance margin over C++ or Rust implementations. The main focus is on the speedup of the algorithms compared to a baseline method, which was chosen to be the well-known Dijkstra's algorithm. All experiments correspond to trajectories on a two-dimensional grid, with variously defined constraints. The motion from each node corresponds to a Moore neighborhood, i.e., it is possible in eight directions. In this paper, three well-known path planning algorithms are described and compared: the Dijkstra, A* and A* /w Bounding Box. And two advanced methods are included, namely Jump Point Search (JPS), incorporated with the Bounding Box variant (JPS+BB), and Simple Subgoal (SS). These advanced methods clearly show their advantage in the context of the speed up of solution time.
{"title":"Explanation and Speedup Comparison of Advanced Path-planning Algorithms Presented on Two-dimensional Grid","authors":"Petr Soustek, R. Matousek, J. Dvorak, Lenka Manakova","doi":"10.13164/mendel.2022.2.097","DOIUrl":"https://doi.org/10.13164/mendel.2022.2.097","url":null,"abstract":"Path planning or network route planning problems are an important issue in AI, robotics, or computer games. Appropriate implementation and knowledge of advanced and classical path-planning algorithms can be important for both autonomous navigation systems and computer games. In this paper, we compare advanced path planning algorithms implemented on a two-dimensional grid. Advanced path planning algorithms, including pseudocode, are introduced. The experiments were performed in the Python environment, thus with a significant performance margin over C++ or Rust implementations. The main focus is on the speedup of the algorithms compared to a baseline method, which was chosen to be the well-known Dijkstra's algorithm. All experiments correspond to trajectories on a two-dimensional grid, with variously defined constraints. The motion from each node corresponds to a Moore neighborhood, i.e., it is possible in eight directions. In this paper, three well-known path planning algorithms are described and compared: the Dijkstra, A* and A* /w Bounding Box. And two advanced methods are included, namely Jump Point Search (JPS), incorporated with the Bounding Box variant (JPS+BB), and Simple Subgoal (SS). These advanced methods clearly show their advantage in the context of the speed up of solution time.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"435 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77407685","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 : 2022-12-20DOI: 10.13164/mendel.2022.2.062
Mhamed-Amine Soumiaa, Sara Elhabbari, Mohamad Mansouri
Cardiovascular diseases (CVD) continues to be the leading cause of death worldwide, with over 17 million deaths each year. In 2015, approximately 422 million people suffered from cardiovascular disease (CVD). Reading and analyzing electrocardiograms (ECGs) can be time consuming, and the development of decision support tools based on automated systems can facilitate and speed up the diagnosis of ECGs. In this paper, we propose a 12 leads ECG signals classification using Multi-level Discrete Wavelet Transform and ResNet34 Deep Learning algorithm which classifies 8 types of cardiovascular diseases: Atrial fibrillation (AF), 1st degree atrioventricular block (AV), Left bundle branch block (LBBB), Right bundle branch block (RBBB), Premature ventricular contraction (PVC), Premature atrial contraction (PAC), ST segment depression (STD), and ST segment elevation (STE). The ECGs are preprocessed, and different features are extracted using Multi-level Discrete Wavelet Transform. The model is trained on a database of more than 6000 electrocardiograms which includes 9 types of 12-lead ECGs: a normal type and the 8 abnormal ones which correspond to the diseases mentioned above.
{"title":"The Use of the Multi-Scale Discrete Wavelet Transform and Deep Neural Networks on ECGs for the Diagnosis of 8 Cardio-Vascular Diseases","authors":"Mhamed-Amine Soumiaa, Sara Elhabbari, Mohamad Mansouri","doi":"10.13164/mendel.2022.2.062","DOIUrl":"https://doi.org/10.13164/mendel.2022.2.062","url":null,"abstract":"Cardiovascular diseases (CVD) continues to be the leading cause of death worldwide, with over 17 million deaths each year. In 2015, approximately 422 million people suffered from cardiovascular disease (CVD). Reading and analyzing electrocardiograms (ECGs) can be time consuming, and the development of decision support tools based on automated systems can facilitate and speed up the diagnosis of ECGs. In this paper, we propose a 12 leads ECG signals classification using Multi-level Discrete Wavelet Transform and ResNet34 Deep Learning algorithm which classifies 8 types of cardiovascular diseases: Atrial fibrillation (AF), 1st degree atrioventricular block (AV), Left bundle branch block (LBBB), Right bundle branch block (RBBB), Premature ventricular contraction (PVC), Premature atrial contraction (PAC), ST segment depression (STD), and ST segment elevation (STE). The ECGs are preprocessed, and different features are extracted using Multi-level Discrete Wavelet Transform. The model is trained on a database of more than 6000 electrocardiograms which includes 9 types of 12-lead ECGs: a normal type and the 8 abnormal ones which correspond to the diseases mentioned above.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"10 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88928768","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 : 2022-12-20DOI: 10.13164/mendel.2022.2.049
Mohammed Y. Kamil, Sahar A. Hashem
Medical imaging, such as chest X-rays, gives an acceptable image of lung functions. Manipulating these images by a radiologist is difficult, thus delaying the diagnosis. Coronavirus is a disease that affects the lung area. Lung segmentation has a significant function in assessing lung disorders. The process of segmentation has seen widespread use of deep learning algorithms. The U-Net is one of the most significant semantic segmentation frameworks for a convolutional neural network. In this paper, the proposed U-Net architecture is evaluated on 565 datasets divided into 500 training images and 65 validation images, For chest X-ray. The findings of the experiments demonstrate that the suggested strategy successfully achieved competitive outcomes with 91.47% and 89.18% accuracy, 0.7494 and 0.7480 IoU, 19.23% and 26.11% loss for training and validation images, respectively.
{"title":"Segmentation of Chest X-Ray Images Using U-Net Model","authors":"Mohammed Y. Kamil, Sahar A. Hashem","doi":"10.13164/mendel.2022.2.049","DOIUrl":"https://doi.org/10.13164/mendel.2022.2.049","url":null,"abstract":"Medical imaging, such as chest X-rays, gives an acceptable image of lung functions. Manipulating these images by a radiologist is difficult, thus delaying the diagnosis. Coronavirus is a disease that affects the lung area. Lung segmentation has a significant function in assessing lung disorders. The process of segmentation has seen widespread use of deep learning algorithms. The U-Net is one of the most significant semantic segmentation frameworks for a convolutional neural network. In this paper, the proposed U-Net architecture is evaluated on 565 datasets divided into 500 training images and 65 validation images, For chest X-ray. The findings of the experiments demonstrate that the suggested strategy successfully achieved competitive outcomes with 91.47% and 89.18% accuracy, 0.7494 and 0.7480 IoU, 19.23% and 26.11% loss for training and validation images, respectively.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"21 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78367494","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 : 2022-06-30DOI: 10.13164/mendel.2022.1.041
G. Kanagaraj, S. S. Sheik Masthan, V. Yu
In robot-assisted manufacturing or assembly, following a predefined path became a critical aspect. In general, inverse kinematics offers the solution to control the movement of manipulator while following the trajectory. The main problem with the inverse kinematics approach is that inverse kinematics are computationally complex. For a redundant manipulator, this complexity is further increased. Instead of employing inverse kinematics, the complexity can be reduced by using a heuristic algorithm. Therefore, a heuristic-based approach can be used to solve the inverse kinematics of the robot manipulator end effector, guaranteeing that the desired paths are accurately followed. This paper compares the performance of four such heuristic-based approaches to solving the inverse kinematics problem. They are Bat Algorithm (BAT), Gravitational Search Algorithm (GSA), Particle Swarm Optimization (PSO), and Whale Optimization Algorithm (WOA). The performance of these algorithms is evaluated based on their ability to accurately follow a predefined trajectory. Extensive simulations show that BAT and GSA outperform PSO and WOA in all aspects considered in this work related to inverse kinematic problems.
{"title":"Meta-Heuristics Based Inverse Kinematics of Robot Manipulator’s Path Tracking Capability Under Joint Limits","authors":"G. Kanagaraj, S. S. Sheik Masthan, V. Yu","doi":"10.13164/mendel.2022.1.041","DOIUrl":"https://doi.org/10.13164/mendel.2022.1.041","url":null,"abstract":"In robot-assisted manufacturing or assembly, following a predefined path became a critical aspect. In general, inverse kinematics offers the solution to control the movement of manipulator while following the trajectory. The main problem with the inverse kinematics approach is that inverse kinematics are computationally complex. For a redundant manipulator, this complexity is further increased. Instead of employing inverse kinematics, the complexity can be reduced by using a heuristic algorithm. Therefore, a heuristic-based approach can be used to solve the inverse kinematics of the robot manipulator end effector, guaranteeing that the desired paths are accurately followed. This paper compares the performance of four such heuristic-based approaches to solving the inverse kinematics problem. They are Bat Algorithm (BAT), Gravitational Search Algorithm (GSA), Particle Swarm Optimization (PSO), and Whale Optimization Algorithm (WOA). The performance of these algorithms is evaluated based on their ability to accurately follow a predefined trajectory. Extensive simulations show that BAT and GSA outperform PSO and WOA in all aspects considered in this work related to inverse kinematic problems.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"49 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77368369","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 : 2022-06-30DOI: 10.13164/mendel.2022.1.001
Rimah Amami, Suleiman Ali Al Saif, Rim Amami, Hassan Ahmed Eleraky, Fatma Melouli, Mariem Baazaoui
The new Coronavirus or simply Covid-19 causes an acute deadly disease. It has spread rapidly across the world, which has caused serious consequences for health professionals and researchers. This is due to many reasons including the lack of vaccine, shortage of testing kits and resources. Therefore, the main purpose of this study is to present an inexpensive alternative diagnostic tool for the detection of Covid-19 infection by using chest radiographs and Deep Convolutional Neural Network (DCNN) technique. In this paper, we have proposed a reliable and economical solution to detect COVID-19. This will be achieved by using X-rays of patients and an Incremental-DCNN (I-DCNN) based on ResNet-101 architecture. The datasets used in this study were collected from publicly available chest radiographs on medical repositories. The proposed I-DCNN method will help in diagnosing the positive Covid-19 patient by utilising three chest X-ray imagery groups, these will be: Covid-19, viral pneumonia, and healthy cases. Furthermore, the main contribution of this paper resides on the use of incremental learning in order to accommodate the detection system. This has high computational energy requirements, time consuming challenges, while working with large-scale and regularly evolving images. The incremental learning process will allow the recognition system to learn new datasets, while keeping the convolutional layers learned previously. The overall Covid-19 detection rate obtained using the proposed I-DCNN was of 98.70% which undeniably can contribute effectively to the detection of COVID-19 infection.
{"title":"The Use of an Incremental Learning Algorithm for Diagnosing COVID-19 from Chest X-ray Images","authors":"Rimah Amami, Suleiman Ali Al Saif, Rim Amami, Hassan Ahmed Eleraky, Fatma Melouli, Mariem Baazaoui","doi":"10.13164/mendel.2022.1.001","DOIUrl":"https://doi.org/10.13164/mendel.2022.1.001","url":null,"abstract":"The new Coronavirus or simply Covid-19 causes an acute deadly disease. It has spread rapidly across the world, which has caused serious consequences for health professionals and researchers. This is due to many reasons including the lack of vaccine, shortage of testing kits and resources. Therefore, the main purpose of this study is to present an inexpensive alternative diagnostic tool for the detection of Covid-19 infection by using chest radiographs and Deep Convolutional Neural Network (DCNN) technique. In this paper, we have proposed a reliable and economical solution to detect COVID-19. This will be achieved by using X-rays of patients and an Incremental-DCNN (I-DCNN) based on ResNet-101 architecture. The datasets used in this study were collected from publicly available chest radiographs on medical repositories. The proposed I-DCNN method will help in diagnosing the positive Covid-19 patient by utilising three chest X-ray imagery groups, these will be: Covid-19, viral pneumonia, and healthy cases. Furthermore, the main contribution of this paper resides on the use of incremental learning in order to accommodate the detection system. This has high computational energy requirements, time consuming challenges, while working with large-scale and regularly evolving images. The incremental learning process will allow the recognition system to learn new datasets, while keeping the convolutional layers learned previously. The overall Covid-19 detection rate obtained using the proposed I-DCNN was of 98.70% which undeniably can contribute effectively to the detection of COVID-19 infection.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"24 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83988645","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 : 2022-06-30DOI: 10.13164/mendel.2022.1.032
Roman Parák, M. Juricek
Advanced robotics does not always have to be associated with Industry 4.0, but can also be applied, for example, in the Smart Hospital concept. Developments in this field have been driven by the coronavirus disease (COVID-19), and any improvement in the work of medical staff is welcome. In this paper, an experimental robotic platform was designed and implemented whose main function is the swabbing samples from the nasal vestibule. The robotic platform represents a complete integration of software and hardware, where the operator has access to a web-based application and can control a number of functions. The increased safety and collaborative approach cannot be overlooked. The result of this work is a functional prototype of the robotic platform that can be further extended, for example, by using alternative technologies, extending patient safety, or clinical tests and studies. Code is available at https://github.com/Steigner/Robo_Medicinae_I
{"title":"Intelligent Sampling of Anterior Human Nasal Swabs using a Collaborative Robotic Arm","authors":"Roman Parák, M. Juricek","doi":"10.13164/mendel.2022.1.032","DOIUrl":"https://doi.org/10.13164/mendel.2022.1.032","url":null,"abstract":"Advanced robotics does not always have to be associated with Industry 4.0, but can also be applied, for example, in the Smart Hospital concept. Developments in this field have been driven by the coronavirus disease (COVID-19), and any improvement in the work of medical staff is welcome. In this paper, an experimental robotic platform was designed and implemented whose main function is the swabbing samples from the nasal vestibule. The robotic platform represents a complete integration of software and hardware, where the operator has access to a web-based application and can control a number of functions. The increased safety and collaborative approach cannot be overlooked. The result of this work is a functional prototype of the robotic platform that can be further extended, for example, by using alternative technologies, extending patient safety, or clinical tests and studies. Code is available at https://github.com/Steigner/Robo_Medicinae_I","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"14 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82448487","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 : 2022-06-30DOI: 10.13164/mendel.2022.1.023
Joao Paulo Schwarz Schuler, S. Romaní, M. Abdel-Nasser, Hatem A. Rashwan, D. Puig
In DCNNs, the number of parameters in pointwise convolutions rapidly grows due to the multiplication of the number of filters by the number of input channels that come from the previous layer. Our proposal makes pointwise convolutions parameter efficient via grouping filters into parallel branches or groups, where each branch processes a fraction of the input channels. However, by doing so, the learning capability of the DCNN is degraded. To avoid this effect, we suggest interleaving the output of filters from different branches at intermediate layers of consecutive pointwise convolutions. We applied our improvement to the EfficientNet, DenseNet-BC L100, MobileNet and MobileNet V3 Large architectures. We trained these architectures with the CIFAR-10, CIFAR-100, Cropped-PlantDoc and The Oxford-IIIT Pet datasets. When training from scratch, we obtained similar test accuracies to the original EfficientNet and MobileNet V3 Large architectures while saving up to 90% of the parameters and 63% of the flops.
{"title":"Grouped Pointwise Convolutions Reduce Parameters in Convolutional Neural Networks","authors":"Joao Paulo Schwarz Schuler, S. Romaní, M. Abdel-Nasser, Hatem A. Rashwan, D. Puig","doi":"10.13164/mendel.2022.1.023","DOIUrl":"https://doi.org/10.13164/mendel.2022.1.023","url":null,"abstract":"In DCNNs, the number of parameters in pointwise convolutions rapidly grows due to the multiplication of the number of filters by the number of input channels that come from the previous layer. Our proposal makes pointwise convolutions parameter efficient via grouping filters into parallel branches or groups, where each branch processes a fraction of the input channels. However, by doing so, the learning capability of the DCNN is degraded. To avoid this effect, we suggest interleaving the output of filters from different branches at intermediate layers of consecutive pointwise convolutions. We applied our improvement to the EfficientNet, DenseNet-BC L100, MobileNet and MobileNet V3 Large architectures. We trained these architectures with the CIFAR-10, CIFAR-100, Cropped-PlantDoc and The Oxford-IIIT Pet datasets. When training from scratch, we obtained similar test accuracies to the original EfficientNet and MobileNet V3 Large architectures while saving up to 90% of the parameters and 63% of the flops.","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76493351","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 : 2022-06-30DOI: 10.13164/mendel.2022.1.063
J. Muller
Advanced optimisation of the aerofoil wing of a general aircraft is the main subject of this paper. Meta-heuristic optimisation techniques, especially swarm algorithms, were used. Subsequently, a new variant denoted as aerofoil particle swarm optimisation (aPSO) was developed from the original particle swarm optimisation (PSO). A parametric model based on B-spline was used to optimise the initial aerofoil. The simulation software Xfoil was calculating basic aerodynamic features (lift, drag, moment).
{"title":"Improving Initial Aerofoil Geometry Using Aerofoil Particle Swarm Optimisation","authors":"J. Muller","doi":"10.13164/mendel.2022.1.063","DOIUrl":"https://doi.org/10.13164/mendel.2022.1.063","url":null,"abstract":"Advanced optimisation of the aerofoil wing of a general aircraft is the main subject of this paper. Meta-heuristic optimisation techniques, especially swarm algorithms, were used. Subsequently, a new variant denoted as aerofoil particle swarm optimisation (aPSO) was developed from the original particle swarm optimisation (PSO). A parametric model based on B-spline was used to optimise the initial aerofoil. The simulation software Xfoil was calculating basic aerodynamic features (lift, drag, moment).","PeriodicalId":38293,"journal":{"name":"Mendel","volume":"43 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84342859","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：