ABSTRACT Protein is always a central part of the biology of the organism, it is essential to be familiar with the nature of proteins’ molecular level communications, in which the prediction of Protein-Protein Interactions (PPIs) plays the main role. This article proposes a new probabilistic feature extraction technique, termed Centroid-based feature (CF) abbreviated as CF-PPI, to generate a new feature from protein sequence, and then the random forest is used as a classifier to predict PPIs. CF-PPI considers the residual energy of the protein bond in the scenario to detect the interaction between proteins and resolve the protein’s length variation issue using probabilistic feature vectors. The PPI datasets which are used in this article are S. cerevisae, H. pylori, and Human, which achieved the average accuracy of 96.25%, 97.68%, and 97.69% respectively using the CF-PPI and Random Forest as a classifier and the comparison result proved superior to other existing results. The AUC score is also evaluated, additionally, a blind test is performed using five other species’ datasets which are independent of the training set with the same proposed feature approach. The experimental results prove that the CF-PPI is very promising and beneficial for looming proteomics research.
{"title":"CF-PPI: Centroid based new feature extraction approach for Protein-Protein Interaction Prediction","authors":"Gunjan Sahni, Bhawna Mewara, Soniya Lalwani, Rajesh Kumar","doi":"10.1080/0952813X.2022.2052189","DOIUrl":"https://doi.org/10.1080/0952813X.2022.2052189","url":null,"abstract":"ABSTRACT Protein is always a central part of the biology of the organism, it is essential to be familiar with the nature of proteins’ molecular level communications, in which the prediction of Protein-Protein Interactions (PPIs) plays the main role. This article proposes a new probabilistic feature extraction technique, termed Centroid-based feature (CF) abbreviated as CF-PPI, to generate a new feature from protein sequence, and then the random forest is used as a classifier to predict PPIs. CF-PPI considers the residual energy of the protein bond in the scenario to detect the interaction between proteins and resolve the protein’s length variation issue using probabilistic feature vectors. The PPI datasets which are used in this article are S. cerevisae, H. pylori, and Human, which achieved the average accuracy of 96.25%, 97.68%, and 97.69% respectively using the CF-PPI and Random Forest as a classifier and the comparison result proved superior to other existing results. The AUC score is also evaluated, additionally, a blind test is performed using five other species’ datasets which are independent of the training set with the same proposed feature approach. The experimental results prove that the CF-PPI is very promising and beneficial for looming proteomics research.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"11 1","pages":"1037 - 1057"},"PeriodicalIF":2.2,"publicationDate":"2022-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73881296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-03-08DOI: 10.1080/0952813X.2022.2046178
Bhanuprakash Dudi, V. Rajesh
ABSTRACT This paper develops a new plant leaf classification model based on enhanced segmentation and optimal feature selection. The first process is the pre-processing, in which RGB to grey-scale conversion, histogram equalisation, and median filtering are adopted. Further, the optimised U-Net model is used for the leaf segmentation. Once the segmentation of the leaf is done, a set of features are extracted related to shape, colour, and texture. Since the length of the feature vector seems to be high that in turn affects the network training, optimal feature selection is adopted in order to reduce data dimensionality and to build robust classification models. Here, the optimal feature selection is performed by the new hybrid algorithm, namely Crow-Electric Fish Optimization (C-EFO), which is the hybridisation of Electric Fish Optimization (EFO) and Crow Search Algorithm (CSA). Finally, the deep learning model termed as Enhanced Recurrent Neural Network (E-RNN) is used for performing the classification with the improvement based on C-EFO. From the analysis, the accuracy of the proposed C-EFO+Opt-U-Net+E-RNN is 4.7% better than k-NN, 3.5% better than VGG16, 3.5% better than LSTM, and 2.75% better than RNN, respectively. Finally, the experimental results on two plant leaf databases show that the proposed method is quite effective and feasible when compared to conventional models.
{"title":"A computer aided plant leaf classification based on optimal feature selection and enhanced recurrent neural network","authors":"Bhanuprakash Dudi, V. Rajesh","doi":"10.1080/0952813X.2022.2046178","DOIUrl":"https://doi.org/10.1080/0952813X.2022.2046178","url":null,"abstract":"ABSTRACT This paper develops a new plant leaf classification model based on enhanced segmentation and optimal feature selection. The first process is the pre-processing, in which RGB to grey-scale conversion, histogram equalisation, and median filtering are adopted. Further, the optimised U-Net model is used for the leaf segmentation. Once the segmentation of the leaf is done, a set of features are extracted related to shape, colour, and texture. Since the length of the feature vector seems to be high that in turn affects the network training, optimal feature selection is adopted in order to reduce data dimensionality and to build robust classification models. Here, the optimal feature selection is performed by the new hybrid algorithm, namely Crow-Electric Fish Optimization (C-EFO), which is the hybridisation of Electric Fish Optimization (EFO) and Crow Search Algorithm (CSA). Finally, the deep learning model termed as Enhanced Recurrent Neural Network (E-RNN) is used for performing the classification with the improvement based on C-EFO. From the analysis, the accuracy of the proposed C-EFO+Opt-U-Net+E-RNN is 4.7% better than k-NN, 3.5% better than VGG16, 3.5% better than LSTM, and 2.75% better than RNN, respectively. Finally, the experimental results on two plant leaf databases show that the proposed method is quite effective and feasible when compared to conventional models.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"43 1","pages":"1001 - 1035"},"PeriodicalIF":2.2,"publicationDate":"2022-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87409382","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-27DOI: 10.1080/0952813X.2022.2035826
P. Varshini, S. Baskar, S. T. Selvi
ABSTRACT A new multiobjective evolutionary optimisation algorithm (MOEA) to solve multimodal, multidimensional, nonconvex, nonlinear, dynamic multiobjective optimisation problems (MOPs) is the need of the hour. The quality of an MOEA lies in a good balance between the exploration and exploitation stages of the MOEA. Utopia constrained MOEA (U-MOEA) is proposed in this paper that improves the exploitation in the replacement step to achieve a perfect balance between exploration and exploitation. The proposed U-MOEA is tested on benchmark MOPs and a multivariable controller design problem. The performance of the proposed algorithm is also compared with other MOEAs such as NSGA-II and ICMDRA concerning hyper volume, nondomination count, combined Pareto set metric, and Cmetric . The performance metrics show better hyper volume and spread metric values for the proposed algorithm, indicating the ability in attaining trade-off region closeness along with diversified Pareto front for U-MOEA when compared to the other two algorithms. Results clearly show that the proposed U-MOEA produces good convergence, diversity characteristics with many numbers of trade-off solutions in a Pareto front.
{"title":"Utopia constrained multi objective optimisation evolutionary algorithm","authors":"P. Varshini, S. Baskar, S. T. Selvi","doi":"10.1080/0952813X.2022.2035826","DOIUrl":"https://doi.org/10.1080/0952813X.2022.2035826","url":null,"abstract":"ABSTRACT A new multiobjective evolutionary optimisation algorithm (MOEA) to solve multimodal, multidimensional, nonconvex, nonlinear, dynamic multiobjective optimisation problems (MOPs) is the need of the hour. The quality of an MOEA lies in a good balance between the exploration and exploitation stages of the MOEA. Utopia constrained MOEA (U-MOEA) is proposed in this paper that improves the exploitation in the replacement step to achieve a perfect balance between exploration and exploitation. The proposed U-MOEA is tested on benchmark MOPs and a multivariable controller design problem. The performance of the proposed algorithm is also compared with other MOEAs such as NSGA-II and ICMDRA concerning hyper volume, nondomination count, combined Pareto set metric, and Cmetric . The performance metrics show better hyper volume and spread metric values for the proposed algorithm, indicating the ability in attaining trade-off region closeness along with diversified Pareto front for U-MOEA when compared to the other two algorithms. Results clearly show that the proposed U-MOEA produces good convergence, diversity characteristics with many numbers of trade-off solutions in a Pareto front.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"19 1","pages":"955 - 971"},"PeriodicalIF":2.2,"publicationDate":"2022-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77974618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-16DOI: 10.1080/0952813X.2021.1960637
Basant Agarwal, A. Agarwal, P. Harjule, Azizur Rahman
ABSTRACT Several studies have been conducted in annotating and collecting the misinformation spread on various social media sites. The misinformation spread during COVID-19 pandemic increased many folds. Understanding the reasons and intent of the misinformation during COVID-19 is a crucial task. Existing approaches have not focused on understanding the intent behind sharing misinformation in the first place. To understand the intent, we introduce a new dataset MisMemoir that apart from annotating misinformation, also collects the social context and site history of the user sharing misinformation. Utilising the established benefits of game theory in social media behaviour analysis, we deploy two-person cooperative games to understand how prominent positive feedback cues like likes and retweets are in motivating an individual to share misinformation on the platform Twitter. Experimental results demonstrate that the spread of misinformation’s primary intent is the intentional/unintentional manoeuvre to increased reach and possibly a false sense of accomplishment. Empirically, we show that in a competitive environment like social media, feedback cues like retweets and comments assume the role of ‘attention’ payoff that significantly affects the strategy of a user on Twitter to share misinformation intentionally.
{"title":"Understanding the intent behind sharing misinformation on social media","authors":"Basant Agarwal, A. Agarwal, P. Harjule, Azizur Rahman","doi":"10.1080/0952813X.2021.1960637","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960637","url":null,"abstract":"ABSTRACT Several studies have been conducted in annotating and collecting the misinformation spread on various social media sites. The misinformation spread during COVID-19 pandemic increased many folds. Understanding the reasons and intent of the misinformation during COVID-19 is a crucial task. Existing approaches have not focused on understanding the intent behind sharing misinformation in the first place. To understand the intent, we introduce a new dataset MisMemoir that apart from annotating misinformation, also collects the social context and site history of the user sharing misinformation. Utilising the established benefits of game theory in social media behaviour analysis, we deploy two-person cooperative games to understand how prominent positive feedback cues like likes and retweets are in motivating an individual to share misinformation on the platform Twitter. Experimental results demonstrate that the spread of misinformation’s primary intent is the intentional/unintentional manoeuvre to increased reach and possibly a false sense of accomplishment. Empirically, we show that in a competitive environment like social media, feedback cues like retweets and comments assume the role of ‘attention’ payoff that significantly affects the strategy of a user on Twitter to share misinformation intentionally.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"12 1","pages":"573 - 587"},"PeriodicalIF":2.2,"publicationDate":"2022-02-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77841441","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-13DOI: 10.1080/0952813X.2021.1960631
A. Parivallal, R. Sakthivel, Chao Wang
ABSTRACT This paper proposes an approach for the solvability of multi-agent systems with Markov jumps subject to time-varying delay and uncertainties. The primary concern of this paper is to construct a state feedback control design with a guaranteed cost function which not only ensures consensus but also assures certain amount of energy consumption. This cost function is designed with the aid of control inputs of all agents and state error among neighbouring agents. A new set of sufficient conditions for the guaranteed cost consensus of the considered Markov jumping multi-agent system (MAS) is derived in terms of linear matrix inequalities (LMIs) by constructing suitable Lyapunov-Krasovskii functional (LKF) and with the aid of Kronecker product properties. Finally, a numerical example is given to illustrate the effectiveness of the developed theoretical results.
{"title":"Guaranteed cost leaderless consensus for uncertain Markov jumping multi-agent systems","authors":"A. Parivallal, R. Sakthivel, Chao Wang","doi":"10.1080/0952813X.2021.1960631","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960631","url":null,"abstract":"ABSTRACT This paper proposes an approach for the solvability of multi-agent systems with Markov jumps subject to time-varying delay and uncertainties. The primary concern of this paper is to construct a state feedback control design with a guaranteed cost function which not only ensures consensus but also assures certain amount of energy consumption. This cost function is designed with the aid of control inputs of all agents and state error among neighbouring agents. A new set of sufficient conditions for the guaranteed cost consensus of the considered Markov jumping multi-agent system (MAS) is derived in terms of linear matrix inequalities (LMIs) by constructing suitable Lyapunov-Krasovskii functional (LKF) and with the aid of Kronecker product properties. Finally, a numerical example is given to illustrate the effectiveness of the developed theoretical results.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"34 6","pages":"257 - 273"},"PeriodicalIF":2.2,"publicationDate":"2022-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72499198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-13DOI: 10.1080/0952813X.2021.1960627
Azam Davahli, M. Shamsi, Golnoush Abaei, Arash Khosravi
ABSTRACT Feature selection (FS) is an optimisation problem that reduces the dimension of the dataset and increases the performance of the machine learning algorithms and classification through the selection of the optimal subset features and elimination of the redundant features. However, the huge search space is an important challenge in the FS problem. Due to their satisfactory capabilities to handle high-dimension search spaces, meta-heuristic search algorithms have recently gained much attention and become popular in the FS problem. For these algorithms, choosing a proper fitness function plays an important role. The fitness function orients the searching strategy of the algorithms to obtain best solutions. Appropriate fitness functions will help the algorithms with exploring the search space more effectively and efficiently. In this work, firstly the efficiency of two of the most outstanding and successful heuristic algorithms in the FS domain, namely genetic algorithm (GA) and grey wolf optimiser (GWO), are investigated and analysed with a single-objective fitness function. Secondly, two recent feature selection techniques based on GA and GWO, namely feature selection, weight, and parameter optimisation (FWP) and binary GWO (BGWO) with their fitness function are investigated and analysed. Thirdly, in order to remove the detected drawbacks and weaknesses of the FS algorithms and to enhance their efficiency, a new multi-objective weighted fitness function based on multiple predominant criteria has been presented. The effectiveness of the proposed fitness function on the FS algorithms is evaluated by using SVM and associative classification on 11 different large and small datasets. The experimental results show the superiority of proposed fitness function (where features were reduced and the classification performance has been improved) over single-objective fitness function and other existing fitness functions. Furthermore, another key aim of this study is to present a comprehensive study about the strengths and weaknesses of the FS algorithms which can be used as guidelines for future possible works to more improve the developments of these algorithms.
特征选择(FS)是一个优化问题，它通过选择最优子集特征和消除冗余特征来降低数据集的维数，提高机器学习算法和分类的性能。然而，巨大的搜索空间是FS问题的一个重要挑战。元启发式搜索算法由于其处理高维搜索空间的令人满意的能力，近年来在FS问题中得到了广泛的关注和应用。在这些算法中，选择合适的适应度函数起着重要的作用。适应度函数指导算法的搜索策略以获得最优解。适当的适应度函数有助于算法更有效地探索搜索空间。本文首先利用单目标适应度函数对遗传算法(genetic algorithm, GA)和灰狼优化器(grey wolf optimizer, GWO)这两种最成功的启发式算法的效率进行了研究和分析。其次，研究了基于遗传算法和GWO的两种最新特征选择技术，即特征选择、权重和参数优化(FWP)和带适应度函数的二元GWO (BGWO)。第三，为了消除现有的多目标加权适应度算法存在的缺陷和不足，提高算法的效率，提出了一种基于多优准则的多目标加权适应度函数。通过在11个不同的大小数据集上使用支持向量机和关联分类来评估所提出的适应度函数对FS算法的有效性。实验结果表明，本文提出的适应度函数(减少了特征，提高了分类性能)优于单目标适应度函数和其他现有适应度函数。此外，本研究的另一个关键目的是对FS算法的优缺点进行全面研究，这可以作为未来可能工作的指导方针，以进一步改进这些算法的发展。
{"title":"Empirical analyses of genetic algorithm and grey wolf optimiser to improve their efficiency with a new multi-objective weighted fitness function for feature selection in machine learning classification: the roadmap","authors":"Azam Davahli, M. Shamsi, Golnoush Abaei, Arash Khosravi","doi":"10.1080/0952813X.2021.1960627","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960627","url":null,"abstract":"ABSTRACT Feature selection (FS) is an optimisation problem that reduces the dimension of the dataset and increases the performance of the machine learning algorithms and classification through the selection of the optimal subset features and elimination of the redundant features. However, the huge search space is an important challenge in the FS problem. Due to their satisfactory capabilities to handle high-dimension search spaces, meta-heuristic search algorithms have recently gained much attention and become popular in the FS problem. For these algorithms, choosing a proper fitness function plays an important role. The fitness function orients the searching strategy of the algorithms to obtain best solutions. Appropriate fitness functions will help the algorithms with exploring the search space more effectively and efficiently. In this work, firstly the efficiency of two of the most outstanding and successful heuristic algorithms in the FS domain, namely genetic algorithm (GA) and grey wolf optimiser (GWO), are investigated and analysed with a single-objective fitness function. Secondly, two recent feature selection techniques based on GA and GWO, namely feature selection, weight, and parameter optimisation (FWP) and binary GWO (BGWO) with their fitness function are investigated and analysed. Thirdly, in order to remove the detected drawbacks and weaknesses of the FS algorithms and to enhance their efficiency, a new multi-objective weighted fitness function based on multiple predominant criteria has been presented. The effectiveness of the proposed fitness function on the FS algorithms is evaluated by using SVM and associative classification on 11 different large and small datasets. The experimental results show the superiority of proposed fitness function (where features were reduced and the classification performance has been improved) over single-objective fitness function and other existing fitness functions. Furthermore, another key aim of this study is to present a comprehensive study about the strengths and weaknesses of the FS algorithms which can be used as guidelines for future possible works to more improve the developments of these algorithms.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"8 1","pages":"171 - 206"},"PeriodicalIF":2.2,"publicationDate":"2022-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84222208","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-13DOI: 10.1080/0952813X.2021.1960629
Zeinab Rahimi Rise, M. Mahootchi, Abbas Ahmadi
ABSTRACT In this paper, we developed an innovative approach combining clustering and classification routines to detect breast cancer severity (stage) and recognise whether or not it metastasises. We use Fuzzy C-mean to cluster data and a proper classification routine to recognise the severity of cancer for each cluster. In other words, we use the divide-and-conquer rule to overcome the nonlinearity of relations between features. Moreover, to have a more accurate classification in the test or real data, we impose the fuzzy membership of each data to a cluster along with other features as the set of input into the classification method. Another advantage of our research study is to use both clinical and image features and to extract new features using principal component analysis (PCA) for the classification phase. Whereas a patient might belong to more than one cluster, the results of all corresponding classification methods for the respective patient are appropriately combined to end up with the stage of the cancerous patient. Ultimately, to investigate the efficiency of the proposed hybrid approach, we use seven real data sets with both clinical and image data.
{"title":"Fusing clinical and image data for detecting the severity of breast cancer by a novel hierarchical approach","authors":"Zeinab Rahimi Rise, M. Mahootchi, Abbas Ahmadi","doi":"10.1080/0952813X.2021.1960629","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960629","url":null,"abstract":"ABSTRACT In this paper, we developed an innovative approach combining clustering and classification routines to detect breast cancer severity (stage) and recognise whether or not it metastasises. We use Fuzzy C-mean to cluster data and a proper classification routine to recognise the severity of cancer for each cluster. In other words, we use the divide-and-conquer rule to overcome the nonlinearity of relations between features. Moreover, to have a more accurate classification in the test or real data, we impose the fuzzy membership of each data to a cluster along with other features as the set of input into the classification method. Another advantage of our research study is to use both clinical and image features and to extract new features using principal component analysis (PCA) for the classification phase. Whereas a patient might belong to more than one cluster, the results of all corresponding classification methods for the respective patient are appropriately combined to end up with the stage of the cancerous patient. Ultimately, to investigate the efficiency of the proposed hybrid approach, we use seven real data sets with both clinical and image data.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"67 1","pages":"207 - 230"},"PeriodicalIF":2.2,"publicationDate":"2022-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86031990","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-13DOI: 10.1080/0952813X.2021.1960630
Saber Gholami, A. Saghiri, S. M. Vahidipour, M. R. Meybodi
ABSTRACT Learning Automata (LAs) are adaptive decision-making models designed to find an appropriate action in unknown environments. LAs can be classified into two classes: variable structure and fixed structure. To the best of our knowledge, there is no hybrid model based on both of these classes. In this paper, we propose a model that brings together the benefits of both classes of LAs. In the proposed model, called an HLA, the action switching phase of a fixed structure learning automaton is fused with a variable structure learning automaton. Several computer simulations are conducted to study the performance of the proposed model with respect to the total number of rewards and action switching in addition to the convergence rate. The proposed model is compared to both variable structure and fixed structure learning automata, and in most cases, the numerical results demonstrate its superiority. In order to show the applicability of the HLA, a novel adaptive dropout mechanism in deep neural networks was suggested. The results of the simulations show that the proposed mechanism performs better than the simple dropout mechanism with respect to network accuracy.
{"title":"HLA: a novel hybrid model based on fixed structure and variable structure learning automata","authors":"Saber Gholami, A. Saghiri, S. M. Vahidipour, M. R. Meybodi","doi":"10.1080/0952813X.2021.1960630","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960630","url":null,"abstract":"ABSTRACT Learning Automata (LAs) are adaptive decision-making models designed to find an appropriate action in unknown environments. LAs can be classified into two classes: variable structure and fixed structure. To the best of our knowledge, there is no hybrid model based on both of these classes. In this paper, we propose a model that brings together the benefits of both classes of LAs. In the proposed model, called an HLA, the action switching phase of a fixed structure learning automaton is fused with a variable structure learning automaton. Several computer simulations are conducted to study the performance of the proposed model with respect to the total number of rewards and action switching in addition to the convergence rate. The proposed model is compared to both variable structure and fixed structure learning automata, and in most cases, the numerical results demonstrate its superiority. In order to show the applicability of the HLA, a novel adaptive dropout mechanism in deep neural networks was suggested. The results of the simulations show that the proposed mechanism performs better than the simple dropout mechanism with respect to network accuracy.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"12 1","pages":"231 - 256"},"PeriodicalIF":2.2,"publicationDate":"2022-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73715465","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-13DOI: 10.1080/0952813X.2021.1960632
R. Suresh, M. Ali, Sumit Saroha
ABSTRACT This paper addresses the global exponential stability in Lagrange sense for memristor-based neural networks (MNNs) with time-varying delays. This paper attempts to derive the delay-dependent Lagrange stability conditions in terms of linear matrix inequalities by designing a suitable Lyapunov-Krasovskii functionaland used Wirtinger inequality, Jensen-based inequality for estimating the integral inequalities. The conditions which are derived confirms the globally exponential stability in Lagrange sense for the proposed MNNs and, the detailed estimation for global exponential attractive set is also given. To show the effectiveness and applicability of the proposed criteria, two numerical examples are also provided in this paper.
{"title":"Global exponential stability of memristor based uncertain neural networks with time-varying delays via Lagrange sense","authors":"R. Suresh, M. Ali, Sumit Saroha","doi":"10.1080/0952813X.2021.1960632","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960632","url":null,"abstract":"ABSTRACT This paper addresses the global exponential stability in Lagrange sense for memristor-based neural networks (MNNs) with time-varying delays. This paper attempts to derive the delay-dependent Lagrange stability conditions in terms of linear matrix inequalities by designing a suitable Lyapunov-Krasovskii functionaland used Wirtinger inequality, Jensen-based inequality for estimating the integral inequalities. The conditions which are derived confirms the globally exponential stability in Lagrange sense for the proposed MNNs and, the detailed estimation for global exponential attractive set is also given. To show the effectiveness and applicability of the proposed criteria, two numerical examples are also provided in this paper.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"8 1","pages":"275 - 288"},"PeriodicalIF":2.2,"publicationDate":"2022-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88816263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-02-13DOI: 10.1080/0952813X.2021.1960626
R. Sakthivel, Karthick S.A, Chao Wang, Kanakalakshmi S
ABSTRACT This paper addresses the reliable finite-time stabilisation problem for a class of fractional-order memristor neural networks under sampled-data controller influenced by the quantisation signal and actuator failures. Precisely, the framework of observer has been initiated for estimating unmeasured state and remunerate the actuator faults with nonlinearities in the controller. Precisely, quantiser is incorporated in the network can reduce the process of transmitting data. Subsequently, activation function approach bringing together with traditional indirect Lyapunov theory endows some sufficient conditions in the frame of linear matrix inequalities to assure the finite-time stabilisation criterion for the addressed neural networks under the proposed reliable sampled-data control. Explicitly, the state feedback control and observer gain matrices are attained by solving the developed linear matrix inequalities. Convincingly, two numerical simulations are explored to substantiate the excellence and potentiality of the developed control law.
{"title":"Finite-time reliable sampled-data control for fractional-order memristive neural networks with quantisation","authors":"R. Sakthivel, Karthick S.A, Chao Wang, Kanakalakshmi S","doi":"10.1080/0952813X.2021.1960626","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1960626","url":null,"abstract":"ABSTRACT This paper addresses the reliable finite-time stabilisation problem for a class of fractional-order memristor neural networks under sampled-data controller influenced by the quantisation signal and actuator failures. Precisely, the framework of observer has been initiated for estimating unmeasured state and remunerate the actuator faults with nonlinearities in the controller. Precisely, quantiser is incorporated in the network can reduce the process of transmitting data. Subsequently, activation function approach bringing together with traditional indirect Lyapunov theory endows some sufficient conditions in the frame of linear matrix inequalities to assure the finite-time stabilisation criterion for the addressed neural networks under the proposed reliable sampled-data control. Explicitly, the state feedback control and observer gain matrices are attained by solving the developed linear matrix inequalities. Convincingly, two numerical simulations are explored to substantiate the excellence and potentiality of the developed control law.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"29 1","pages":"109 - 127"},"PeriodicalIF":2.2,"publicationDate":"2022-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72486611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","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：