Vision-language models (VLMs), pretrained on expansive datasets containing image-text pairs, have exhibited remarkable transferability across a diverse spectrum of visual tasks. The leveraging of knowledge encoded within these potent VLMs holds significant promise for the advancement of effective video recognition models. A fundamental aspect of pretrained VLMs lies in their ability to establish a crucial bridge between the visual and textual domains. In our pioneering work, we introduce the Attr4Vis framework, dedicated to exploring knowledge transfer between Video and Text modalities to bolster video recognition performance. Central to our contributions is the comprehensive revisitation of Text-to-Video classifier initialization, a critical step that refines the initialization process and streamlines the integration of our framework, particularly within existing Vision-Language Models (VLMs). Furthermore, we emphasize the adoption of dense attribute generation techniques, shedding light on their paramount importance in video analysis. By effectively encoding attribute changes over time, these techniques significantly enhance event representation and recognition within videos. In addition, we introduce an innovative Attribute Enrichment Algorithm aimed at enriching set of attributes by large language models (LLMs) like ChatGPT. Through the seamless integration of these components, Attr4Vis attains a state-of-the-art accuracy of 91.5% on the challenging Kinetics-400 dataset using the InternVideo model.
{"title":"Attr4Vis: Revisiting Importance of Attribute Classification in Vision-Language Models for Video Recognition","authors":"Alexander Zarichkovyi, Inna V. Stetsenko","doi":"10.47839/ijc.23.1.3440","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3440","url":null,"abstract":"Vision-language models (VLMs), pretrained on expansive datasets containing image-text pairs, have exhibited remarkable transferability across a diverse spectrum of visual tasks. The leveraging of knowledge encoded within these potent VLMs holds significant promise for the advancement of effective video recognition models. A fundamental aspect of pretrained VLMs lies in their ability to establish a crucial bridge between the visual and textual domains. In our pioneering work, we introduce the Attr4Vis framework, dedicated to exploring knowledge transfer between Video and Text modalities to bolster video recognition performance. Central to our contributions is the comprehensive revisitation of Text-to-Video classifier initialization, a critical step that refines the initialization process and streamlines the integration of our framework, particularly within existing Vision-Language Models (VLMs). Furthermore, we emphasize the adoption of dense attribute generation techniques, shedding light on their paramount importance in video analysis. By effectively encoding attribute changes over time, these techniques significantly enhance event representation and recognition within videos. In addition, we introduce an innovative Attribute Enrichment Algorithm aimed at enriching set of attributes by large language models (LLMs) like ChatGPT. Through the seamless integration of these components, Attr4Vis attains a state-of-the-art accuracy of 91.5% on the challenging Kinetics-400 dataset using the InternVideo model.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140760739","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}
The article discusses the method of identifying parameters for interval nonlinear models of static systems. The method is based on solving an optimization problem with a smooth objective function. Additional coefficients are added to the objective function's variables to solve the optimization problem, complicating the computational procedures. The computational complexity of quasi-Newton methods used to solve the optimization problem is analyzed. Excessive computational complexity is caused by many iterations when transforming the value of the objective function to zero. To address this, the article proposes using the optimization stop criterion based on the determination of the model's adequacy at the current iteration of the computational optimization procedure. Numerical experiments were conducted to identify nonlinear models of depending the pH of the environment in the fermenter of the biogas plant on influencing factors. It was established that the proposed criterion reduced the number of iterations by 4.5 times, which is proportional to the same reduction in the number of calculations of the objective function. Gotten results are also important for reducing the computational complexity of algorithms of structural identification of these models.
{"title":"The Improved Method for Identifying Parameters of Interval Nonlinear Models of Static Systems","authors":"Volodymyr Manzhula, M. Dyvak, Vadym Zabchuk","doi":"10.47839/ijc.23.1.3431","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3431","url":null,"abstract":"The article discusses the method of identifying parameters for interval nonlinear models of static systems. The method is based on solving an optimization problem with a smooth objective function. Additional coefficients are added to the objective function's variables to solve the optimization problem, complicating the computational procedures. The computational complexity of quasi-Newton methods used to solve the optimization problem is analyzed. Excessive computational complexity is caused by many iterations when transforming the value of the objective function to zero. To address this, the article proposes using the optimization stop criterion based on the determination of the model's adequacy at the current iteration of the computational optimization procedure. Numerical experiments were conducted to identify nonlinear models of depending the pH of the environment in the fermenter of the biogas plant on influencing factors. It was established that the proposed criterion reduced the number of iterations by 4.5 times, which is proportional to the same reduction in the number of calculations of the objective function. Gotten results are also important for reducing the computational complexity of algorithms of structural identification of these models.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140771573","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}
The detection of human emotions from speech signals remains a challenging frontier in audio processing and human-computer interaction domains. This study introduces a novel approach to Speech Emotion Recognition (SER) using a Dendritic Layer combined with a Capsule Network (DendCaps). A Convolutional Neural Network (NN) and a Long Short-Time Neural Network (CLSTM) hybrid model are used to create a baseline which is then compared to the DendCap model. Integrating dendritic layers and capsule networks for speech emotion detection can harness the unique advantages of both architectures, potentially leading to more sophisticated and accurate models. Dendritic layers, inspired by the nonlinear processing properties of dendritic trees in biological neurons, can handle the intricate patterns and variabilities inherent in speech signals, while capsule networks, with their dynamic routing mechanisms, are adept at preserving hierarchical spatial relationships within the data, enabling the model to capture more refined emotional subtleties in human speech. The main motivation for using DendCaps is to bridge the gap between the capabilities of biological neural systems and artificial neural networks. This combination aims to capitalize on the hierarchical nature of speech data, where intricate patterns and dependencies can be better captured. Finally, two ensemble methods namely stacking and boosting are used for evaluating the CLSTM and DendCaps networks and the experimental results show that stacking of the CLSTM and DendCaps networks gives the superior result with a 75% accuracy.
{"title":"Speech Emotion Recognition using Hybrid Architectures","authors":"Michael Norval, Zenghui Wang","doi":"10.47839/ijc.23.1.3430","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3430","url":null,"abstract":"The detection of human emotions from speech signals remains a challenging frontier in audio processing and human-computer interaction domains. This study introduces a novel approach to Speech Emotion Recognition (SER) using a Dendritic Layer combined with a Capsule Network (DendCaps). A Convolutional Neural Network (NN) and a Long Short-Time Neural Network (CLSTM) hybrid model are used to create a baseline which is then compared to the DendCap model. Integrating dendritic layers and capsule networks for speech emotion detection can harness the unique advantages of both architectures, potentially leading to more sophisticated and accurate models. Dendritic layers, inspired by the nonlinear processing properties of dendritic trees in biological neurons, can handle the intricate patterns and variabilities inherent in speech signals, while capsule networks, with their dynamic routing mechanisms, are adept at preserving hierarchical spatial relationships within the data, enabling the model to capture more refined emotional subtleties in human speech. The main motivation for using DendCaps is to bridge the gap between the capabilities of biological neural systems and artificial neural networks. This combination aims to capitalize on the hierarchical nature of speech data, where intricate patterns and dependencies can be better captured. Finally, two ensemble methods namely stacking and boosting are used for evaluating the CLSTM and DendCaps networks and the experimental results show that stacking of the CLSTM and DendCaps networks gives the superior result with a 75% accuracy.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140784957","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}
This study aims to evaluate the quality of the Region III Higher Education Service Institution (RHSI3) website using the modified WebQual 4.0. This evaluation needs to be carried out to find out things that need to be improved on the website so that it can satisfy its users. Based on the evaluation results, it can be seen that the overall average score of the RHSI3 website measurement is 568.50 with an interpretation of 69.33%, so it is included in the Good criteria. There are six indicators that get a score above the average while there are four indicators that get a score below the average. The indicator that gets the highest score is the indicator about the simplicity of learning to operate the RHSI3 website. This indicator gets a score of 808 with an interpretation of 98.58%, so it is included in the Excellent criteria. The indicator that gets the lowest score is whether the website provides a space for the community. The indicator only gets a score of 216 with an interpretation of 26.34%, so it is included in the Bad criteria. To improve the quality of the website, it is necessary to improve several indicators that get a low score interpretation value, i.e., providing detailed information, a space for the community, and making it easier to communicate with organizations. From an academic point of view, this study contributes to the modifications of WebQual 4.0 as well as gives examples of how to use it. From a practical point of view, the results of this study can be a reference for RHSI3 website managers regarding things that need to be considered and improved to make their website quality better.
{"title":"Website Quality Measurement of Educational Government Agency in Indonesia using Modified WebQual 4.0","authors":"Rintho Rante Rerung, Arief Ramadhan","doi":"10.47839/ijc.23.1.3432","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3432","url":null,"abstract":"This study aims to evaluate the quality of the Region III Higher Education Service Institution (RHSI3) website using the modified WebQual 4.0. This evaluation needs to be carried out to find out things that need to be improved on the website so that it can satisfy its users. Based on the evaluation results, it can be seen that the overall average score of the RHSI3 website measurement is 568.50 with an interpretation of 69.33%, so it is included in the Good criteria. There are six indicators that get a score above the average while there are four indicators that get a score below the average. The indicator that gets the highest score is the indicator about the simplicity of learning to operate the RHSI3 website. This indicator gets a score of 808 with an interpretation of 98.58%, so it is included in the Excellent criteria. The indicator that gets the lowest score is whether the website provides a space for the community. The indicator only gets a score of 216 with an interpretation of 26.34%, so it is included in the Bad criteria. To improve the quality of the website, it is necessary to improve several indicators that get a low score interpretation value, i.e., providing detailed information, a space for the community, and making it easier to communicate with organizations. From an academic point of view, this study contributes to the modifications of WebQual 4.0 as well as gives examples of how to use it. From a practical point of view, the results of this study can be a reference for RHSI3 website managers regarding things that need to be considered and improved to make their website quality better.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140757236","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}
Ayman Alsabry, Malek Algabri, Amin Mohamed Ahsan, M. A. Mosleh, F. E. Hanash, Hamzah Ali Qasem
Breast cancer is a primary cause of cancer-associated mortality among women globally, and early detection and personalized treatment are critical for improving patient outcomes. In this study, we propose an optimal framework for predicting breast cancer patient survivability using the GentleBoost algorithm and Bayesian optimization. The proposed framework combines the strengths of the GentleBoost algorithm, which is a powerful machine-learning algorithm for classification, and Bayesian optimization, which is a powerful optimization technique for hyperparameter tuning. We evaluated the proposed framework using the publicly available breast cancer dataset provided by The Surveillance, Epidemiology, and End Results (SEER) program and compared its performance with several popular single algorithms, including support vector machine (SVM), artificial neural network (ANN), and k-nearest neighbors (KNN). The experimental results demonstrate that the proposed framework outperforms these methods in terms of accuracy (mean= 95.16%, best = 95.35, worst = 95.1%, and SD = 0.008). The values of precision, recall, and f1-score of the best experiment were 92.3 %, 98.2 %, and 95.2 %, respectively, with hyperparameters of (number of learners = 246, learning rate = 0.0011, and maximum number of splits = 1240). The proposed framework has the potential to improve breast cancer patient survival predictions and personalized treatment plans, leading to the improved patient outcomes and reduced healthcare costs.
{"title":"An Optimal Framework Based on the GentleBoost Algorithm and Bayesian Optimization for the Prediction of Breast Cancer Patients' Survivability","authors":"Ayman Alsabry, Malek Algabri, Amin Mohamed Ahsan, M. A. Mosleh, F. E. Hanash, Hamzah Ali Qasem","doi":"10.47839/ijc.23.1.3439","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3439","url":null,"abstract":"Breast cancer is a primary cause of cancer-associated mortality among women globally, and early detection and personalized treatment are critical for improving patient outcomes. In this study, we propose an optimal framework for predicting breast cancer patient survivability using the GentleBoost algorithm and Bayesian optimization. The proposed framework combines the strengths of the GentleBoost algorithm, which is a powerful machine-learning algorithm for classification, and Bayesian optimization, which is a powerful optimization technique for hyperparameter tuning. We evaluated the proposed framework using the publicly available breast cancer dataset provided by The Surveillance, Epidemiology, and End Results (SEER) program and compared its performance with several popular single algorithms, including support vector machine (SVM), artificial neural network (ANN), and k-nearest neighbors (KNN). The experimental results demonstrate that the proposed framework outperforms these methods in terms of accuracy (mean= 95.16%, best = 95.35, worst = 95.1%, and SD = 0.008). The values of precision, recall, and f1-score of the best experiment were 92.3 %, 98.2 %, and 95.2 %, respectively, with hyperparameters of (number of learners = 246, learning rate = 0.0011, and maximum number of splits = 1240). The proposed framework has the potential to improve breast cancer patient survival predictions and personalized treatment plans, leading to the improved patient outcomes and reduced healthcare costs.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140783474","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}
The paper considers the speedy images processing in Wireless Multimedia Sensor Networks using the Residue Number System (RNS) and the method of arithmetic coding. The proposed method has a two-stage frame: firstly, the RNS transformation is run to divide the data and obtain residues, and secondly, the parallel compression of the resulting residues is provided by employing the arithmetic coding. Within the implementation of binary code transformation in RNS one, the hardware complexity for block conversion is evaluated for various modulo sets and the results are illustrated. Authors employed the arithmetic coding for residue compression to provide the optimum of compression degree in terms of entropy assessment as well as a reduction in image redundancy without loss of quality. A research algorithm is proposed to run an experiment presented by the residues carried out on test images and other types of files. As a result, an increase in the speed of image compression of about 2.5 times is achieved by processing the small data as well as providing the parallel operation of the compression residue units by RNS selected moduli. Finally, the existing and proposed methods are compared and it has been shown the last one provides a better compression ratio of more than twice.
{"title":"Image Transmission in WMSN Based on Residue Number System","authors":"A. Sachenko, V. Yatskiv, Jürgen Sieck, Jun Su","doi":"10.47839/ijc.23.1.3444","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3444","url":null,"abstract":"The paper considers the speedy images processing in Wireless Multimedia Sensor Networks using the Residue Number System (RNS) and the method of arithmetic coding. The proposed method has a two-stage frame: firstly, the RNS transformation is run to divide the data and obtain residues, and secondly, the parallel compression of the resulting residues is provided by employing the arithmetic coding. Within the implementation of binary code transformation in RNS one, the hardware complexity for block conversion is evaluated for various modulo sets and the results are illustrated. Authors employed the arithmetic coding for residue compression to provide the optimum of compression degree in terms of entropy assessment as well as a reduction in image redundancy without loss of quality. A research algorithm is proposed to run an experiment presented by the residues carried out on test images and other types of files. As a result, an increase in the speed of image compression of about 2.5 times is achieved by processing the small data as well as providing the parallel operation of the compression residue units by RNS selected moduli. Finally, the existing and proposed methods are compared and it has been shown the last one provides a better compression ratio of more than twice.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140776265","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}
This article presents a comparative study of two validation approaches in .NET – Data Annotations and Fluent Validation – analyzing their syntax, functionality, and other factors (such as readability, maintainability, and performance). The study begins by examining the Data Annotations approach, an in-built validation mechanism in the .NET Framework that uses validation attributes to validate model properties. While Data Annotations offers a simple syntax and is well-known to .NET developers, it may not be ideal for more complex validation scenarios and could become verbose and difficult to maintain. The study then introduces the Fluent Validation approach, which utilizes a fluent syntax to define validation rules in a more expressive, readable, and concise manner. With its flexible architecture and fluent API (application programming interface), Fluent Validation provides greater control over the validation process, enabling better maintainability and performance. The study concludes by highlighting the merits and drawbacks of both approaches, noting that the choice of validation approach will depend on the specific requirements of the project at hand.
{"title":"A Comparative Study of Data Annotations and Fluent Validation in .NET","authors":"V. Samotyy, U. Dzelendzyak, N. Mashtaler","doi":"10.47839/ijc.23.1.3437","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3437","url":null,"abstract":"This article presents a comparative study of two validation approaches in .NET – Data Annotations and Fluent Validation – analyzing their syntax, functionality, and other factors (such as readability, maintainability, and performance). The study begins by examining the Data Annotations approach, an in-built validation mechanism in the .NET Framework that uses validation attributes to validate model properties. While Data Annotations offers a simple syntax and is well-known to .NET developers, it may not be ideal for more complex validation scenarios and could become verbose and difficult to maintain. The study then introduces the Fluent Validation approach, which utilizes a fluent syntax to define validation rules in a more expressive, readable, and concise manner. With its flexible architecture and fluent API (application programming interface), Fluent Validation provides greater control over the validation process, enabling better maintainability and performance. The study concludes by highlighting the merits and drawbacks of both approaches, noting that the choice of validation approach will depend on the specific requirements of the project at hand.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140759350","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}
Scheduling workflow tasks in grid computing is a complex process, especially if it is associated with satisfying the user's requirements to complete tasks within a specified time, with lowest possible cost. This paper presents a proposed Simulated Annealing (SA) based Grid Workflow Tasks Scheduling Approach (SA-GWTSA) that takes into account users’ QoS (quality of service) constraints in terms of cost and time. For a given set of inter-dependent workflow tasks, it generates an optimal schedule, which minimizes the execution time and cost, such that the optimized time is within the time constraints (deadline) imposed by the user. In SA-GWTSA, the workflow tasks, which are modeled as a DAG, are divided into task divisions, each of which consists of a set of sequential tasks. Then, the optimal sub-schedules of all task divisions are computed applying SA algorithm, and used to obtain the execution schedule of the entire workflow. In the proposed algorithm, the sub-schedule of each branch division is represented by a vector, in which each element holds the ID of the service provider chosen from a list of service providers capable of executing the corresponding task in the branch. The algorithm uses a fitness function that is formulated as a multi-objective function of time and cost, which gives users the ability to determine their requirements of time against cost, by changing the weighting coefficients in the objective function. The paper also exhibits the experimental results of assessing the performance of SA-GWTSA with workflows samples of different sizes, compared to different scheduling algorithms: Greedy-Time, Greedy-Cost, and Modified Greedy-Cost.
{"title":"SA-Based QoS Aware Workflow Scheduling of Collaborative Tasks in Grid Computing","authors":"M. Girgis, Tarek M. Mahmoud, Hagar M. Azzam","doi":"10.47839/ijc.23.1.3436","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3436","url":null,"abstract":"Scheduling workflow tasks in grid computing is a complex process, especially if it is associated with satisfying the user's requirements to complete tasks within a specified time, with lowest possible cost. This paper presents a proposed Simulated Annealing (SA) based Grid Workflow Tasks Scheduling Approach (SA-GWTSA) that takes into account users’ QoS (quality of service) constraints in terms of cost and time. For a given set of inter-dependent workflow tasks, it generates an optimal schedule, which minimizes the execution time and cost, such that the optimized time is within the time constraints (deadline) imposed by the user. In SA-GWTSA, the workflow tasks, which are modeled as a DAG, are divided into task divisions, each of which consists of a set of sequential tasks. Then, the optimal sub-schedules of all task divisions are computed applying SA algorithm, and used to obtain the execution schedule of the entire workflow. In the proposed algorithm, the sub-schedule of each branch division is represented by a vector, in which each element holds the ID of the service provider chosen from a list of service providers capable of executing the corresponding task in the branch. The algorithm uses a fitness function that is formulated as a multi-objective function of time and cost, which gives users the ability to determine their requirements of time against cost, by changing the weighting coefficients in the objective function. The paper also exhibits the experimental results of assessing the performance of SA-GWTSA with workflows samples of different sizes, compared to different scheduling algorithms: Greedy-Time, Greedy-Cost, and Modified Greedy-Cost.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140788260","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}
Vadym I. Slusar, Andrii Zinchenko, Yuriy Danyk, M. Klymash, Yuliia Pyrih
The method of joint processing of pulses and OFDM (N-OFDM) signals is proposed. The corresponding analytical relations for the lower Cramer-Rao boundary on the dispersion of OFDM (N-OFDM) signals amplitude ratings in the presence of sources of pulsed radiation are obtained. Using mathematical modeling properties and limitations of the demodulation method of OFDM (N-OFDM) signals in the background of impulse signals in the integrated radar and telecommunication systems are established. It is determined that the use of the angular distance between the pulsed and OFDM signals sources at a value that is not less than 0.75 widths of the secondary beam of the digital antenna array pattern does not affect the accuracy of the OFDM signal amplitudes. The same applies to the active interferences.
{"title":"Spatial Selection-Based Intelligent N-OFDM Signal Processing in Wireless Communication Systems","authors":"Vadym I. Slusar, Andrii Zinchenko, Yuriy Danyk, M. Klymash, Yuliia Pyrih","doi":"10.47839/ijc.23.1.3433","DOIUrl":"https://doi.org/10.47839/ijc.23.1.3433","url":null,"abstract":"The method of joint processing of pulses and OFDM (N-OFDM) signals is proposed. The corresponding analytical relations for the lower Cramer-Rao boundary on the dispersion of OFDM (N-OFDM) signals amplitude ratings in the presence of sources of pulsed radiation are obtained. Using mathematical modeling properties and limitations of the demodulation method of OFDM (N-OFDM) signals in the background of impulse signals in the integrated radar and telecommunication systems are established. It is determined that the use of the angular distance between the pulsed and OFDM signals sources at a value that is not less than 0.75 widths of the secondary beam of the digital antenna array pattern does not affect the accuracy of the OFDM signal amplitudes. The same applies to the active interferences.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140797134","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}
This paper introduces Learnable Extended Activation Function (LEAF) - an adaptive activation function that combines the properties of squashing functions and rectifier units. Depending on the target architecture and data processing task, LEAF adapts its form during training to achieve lower loss values and improve the training results. While not suffering from the "vanishing gradient" effect, LEAF can directly replace SiLU, ReLU, Sigmoid, Tanh, Swish, and AHAF in feed-forward, recurrent, and many other neural network architectures. The training process for LEAF features a two-stage approach when the activation function parameters update before the synaptic weights. The experimental evaluation in the image classification task shows the superior performance of LEAF compared to the non-adaptive alternatives. Particularly, LEAF-asTanh provides 7% better classification accuracy than hyperbolic tangents on the CIFAR-10 dataset. As empirically examined, LEAF-as-SiLU and LEAF-as-Sigmoid in convolutional networks tend to "evolve" into SiLU-like forms. The proposed activation function and the corresponding training algorithm are relatively simple from the computational standpoint and easily apply to existing deep neural networks.
{"title":"Learnable Extended Activation Function for Deep Neural Networks","authors":"Yevgeniy Bodyanskiy, Serhii Kostiuk","doi":"10.47839/ijc.22.3.3225","DOIUrl":"https://doi.org/10.47839/ijc.22.3.3225","url":null,"abstract":"This paper introduces Learnable Extended Activation Function (LEAF) - an adaptive activation function that combines the properties of squashing functions and rectifier units. Depending on the target architecture and data processing task, LEAF adapts its form during training to achieve lower loss values and improve the training results. While not suffering from the \"vanishing gradient\" effect, LEAF can directly replace SiLU, ReLU, Sigmoid, Tanh, Swish, and AHAF in feed-forward, recurrent, and many other neural network architectures. The training process for LEAF features a two-stage approach when the activation function parameters update before the synaptic weights. The experimental evaluation in the image classification task shows the superior performance of LEAF compared to the non-adaptive alternatives. Particularly, LEAF-asTanh provides 7% better classification accuracy than hyperbolic tangents on the CIFAR-10 dataset. As empirically examined, LEAF-as-SiLU and LEAF-as-Sigmoid in convolutional networks tend to \"evolve\" into SiLU-like forms. The proposed activation function and the corresponding training algorithm are relatively simple from the computational standpoint and easily apply to existing deep neural networks.","PeriodicalId":37669,"journal":{"name":"International Journal of Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135458235","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：