In India, the Public Distribution System (PDS) is a critical tool for accomplishing the aim of “Zero Hunger”. Despite the enormous resources used, PDS has several inefficiencies that are caused by the monopoly of agents engaged in last-mile grain supply. Various state governments in India have been employing portability as an innovative solution to address this problem. In this article, we examined a huge-scale data on the deployment of portable beneficiaries arriving in a particular FPS of Kerala state in India over three years. A comparison is made between Auto-Regressive Integrated Moving Average (ARIMA) method which makes forecasts in univariate data and ARIMA with exogenous variables called ARIMAX. We followed Mean Absolute Percentage Error (MAPE) and Mean Absolute Deviation (MAD) as the accuracy performance measure of the models and observed that the ARIMAX model outperforms the ARIMA model with the least forecasting errors.
{"title":"Forecasting stochastic consumer portability visitation pattern in fair price shops of India","authors":"A. Sasi, Thiruselvan Subramanian","doi":"10.47974/jios-1364","DOIUrl":"https://doi.org/10.47974/jios-1364","url":null,"abstract":"In India, the Public Distribution System (PDS) is a critical tool for accomplishing the aim of “Zero Hunger”. Despite the enormous resources used, PDS has several inefficiencies that are caused by the monopoly of agents engaged in last-mile grain supply. Various state governments in India have been employing portability as an innovative solution to address this problem. In this article, we examined a huge-scale data on the deployment of portable beneficiaries arriving in a particular FPS of Kerala state in India over three years. A comparison is made between Auto-Regressive Integrated Moving Average (ARIMA) method which makes forecasts in univariate data and ARIMA with exogenous variables called ARIMAX. We followed Mean Absolute Percentage Error (MAPE) and Mean Absolute Deviation (MAD) as the accuracy performance measure of the models and observed that the ARIMAX model outperforms the ARIMA model with the least forecasting errors.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"1 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"70470551","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}
N. Madhavi, Sushil Dohare, G. Prasad, D. Babu, Abdul Rahman Mohammed Al-Ansari
According to the world health organization, every year, more than 8% of women suffer due to breast cancer, and 40% of women die in low-poverty regions. This entire work focuses on the algorithm to detect breast cancer. This algorithm improves the accuracy of the detection and the risk factor determination by using the backpropagation network (BPN) theory and the Support vector method (SVM). By the end of the entire work, the improved accuracy is up to 95% compared to other forms; this proposed method is proper when evaluating the patient report in the image format, like a scanning report.
{"title":"Enhancing the accuracy of breast cancer detection and determination of risk factor by using the backpropagation network theory and SVM: Machine learning","authors":"N. Madhavi, Sushil Dohare, G. Prasad, D. Babu, Abdul Rahman Mohammed Al-Ansari","doi":"10.47974/jios-1365","DOIUrl":"https://doi.org/10.47974/jios-1365","url":null,"abstract":"According to the world health organization, every year, more than 8% of women suffer due to breast cancer, and 40% of women die in low-poverty regions. This entire work focuses on the algorithm to detect breast cancer. This algorithm improves the accuracy of the detection and the risk factor determination by using the backpropagation network (BPN) theory and the Support vector method (SVM). By the end of the entire work, the improved accuracy is up to 95% compared to other forms; this proposed method is proper when evaluating the patient report in the image format, like a scanning report.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"7 1","pages":""},"PeriodicalIF":1.4,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"70470593","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 : 2022-11-17DOI: 10.1080/02522667.2022.2133215
A. Khobragade, Rushikesh Mahajan, Hrithik Langi, Rohit Mundhe, S. Ghumbre
Abstract Knowledge graphs contain only positive triplet facts, whereas the negative triplets need to be generated precisely to train the embedding models. Early Uniform and Bernoulli sampling are applied but suffer’s from the zero loss problems during training, affecting the performance of embedding models. Recently, generative adversarial technic attended the dynamic negative sampling and obtained better performance by vanishing zero loss but on the adverse side of increasing the model complexity and training parameter. However, NSCaching balances the performance and complexity, generating a single negative triplet sample for each positive triplet that focuses on vanishing gradients. This paper addressed the zero loss training problem due to the low-scored negative triplet by proposing the extended version of NSCaching, to generate the high-scored negative triplet utilized to increase the training performance. The proposed method experimented with semantic matching knowledge graph embedding models on the benchmark datasets, where the results show the success on all evaluation metrics.
{"title":"Effective negative triplet sampling for knowledge graph embedding","authors":"A. Khobragade, Rushikesh Mahajan, Hrithik Langi, Rohit Mundhe, S. Ghumbre","doi":"10.1080/02522667.2022.2133215","DOIUrl":"https://doi.org/10.1080/02522667.2022.2133215","url":null,"abstract":"Abstract Knowledge graphs contain only positive triplet facts, whereas the negative triplets need to be generated precisely to train the embedding models. Early Uniform and Bernoulli sampling are applied but suffer’s from the zero loss problems during training, affecting the performance of embedding models. Recently, generative adversarial technic attended the dynamic negative sampling and obtained better performance by vanishing zero loss but on the adverse side of increasing the model complexity and training parameter. However, NSCaching balances the performance and complexity, generating a single negative triplet sample for each positive triplet that focuses on vanishing gradients. This paper addressed the zero loss training problem due to the low-scored negative triplet by proposing the extended version of NSCaching, to generate the high-scored negative triplet utilized to increase the training performance. The proposed method experimented with semantic matching knowledge graph embedding models on the benchmark datasets, where the results show the success on all evaluation metrics.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2075 - 2087"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45446962","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 : 2022-11-17DOI: 10.1080/02522667.2022.2133218
D. Kshirsagar, Pooja Agrawal
Abstract Feature Selection (FS) provides a vital role in the android malware detection system. The researchers have presented FS methods and tested them on benchmark datasets, including static types of features extracted from applications. This paper studies FS methods used in traditional android malware detection systems. These FS methods are implemented on the benchmark datasets such as Genome project, Google PlayStore, AndroZoo, and Drebin consist of static types of extracted features from applications. These traditional methods are studied and implemented on the latest dataset, such as CIC-MalDroid2020 dataset, which includes the latest types of malware and 470 dynamic types of features. The experimentation is performed on CIC-MalDroid2020 dataset with the Random Forest (RF) classifier using traditional FS methods, and performance is compared with the original feature set. Finally, the investigation with the RF classifier on CIC-MalDroid2020 dataset using the obtained 80 features from 470 original features by the ReliefF method produces higher precision of 97.4647% and a lower FAR of 0.1409 for malware detection.
{"title":"A study of feature selection methods for android malware detection","authors":"D. Kshirsagar, Pooja Agrawal","doi":"10.1080/02522667.2022.2133218","DOIUrl":"https://doi.org/10.1080/02522667.2022.2133218","url":null,"abstract":"Abstract Feature Selection (FS) provides a vital role in the android malware detection system. The researchers have presented FS methods and tested them on benchmark datasets, including static types of features extracted from applications. This paper studies FS methods used in traditional android malware detection systems. These FS methods are implemented on the benchmark datasets such as Genome project, Google PlayStore, AndroZoo, and Drebin consist of static types of extracted features from applications. These traditional methods are studied and implemented on the latest dataset, such as CIC-MalDroid2020 dataset, which includes the latest types of malware and 470 dynamic types of features. The experimentation is performed on CIC-MalDroid2020 dataset with the Random Forest (RF) classifier using traditional FS methods, and performance is compared with the original feature set. Finally, the investigation with the RF classifier on CIC-MalDroid2020 dataset using the obtained 80 features from 470 original features by the ReliefF method produces higher precision of 97.4647% and a lower FAR of 0.1409 for malware detection.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2111 - 2120"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47692180","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 : 2022-11-17DOI: 10.1080/02522667.2022.2133222
Sakshi Dhingra, Dharminder Kumar
Abstract Hyperspectral images usually comprise several continuous spectral bands that represent the category of similar objects or material within the captured scene. These high-dimensional data structures have a high level of correlation and possess unique information that can be used for precise image classification. The precise selection of useful features from these high dimensional band information is very important to reduce the challenge of hyper spectral image classification approaches. Nowadays, metaheuristic algorithms are immensely utilized as a promising tool for hyperspectral image classification. In the present research work, hyperspectral images are classified with the various combinations of meta-heuristic approaches and the neural network including the mostly used Cuckoo Search (CS) optimization algorithm to resolve the global optimization search problems considering the improvement needed in image classification. Further, the strength of CS is improved using the integration of the Genetic Algorithm (GA) fitness function within the CS. The feature selection is performed by the hybrid CS and GA algorithm and the optimized features are then fed to ANN for training and classification. The paper has shown a comparative analysis of various meta heuristics techniques with ANN on parameters like kappa coefficient, Class accuracy and overall Accuracy and the designed algorithms are tested on the Indian Pines dataset. The proposed CS and GA with ANN outperformed the two already existing works with an overall average accuracy of 97.30% and a kappa coefficient of 0.9760.
{"title":"Hyperspectral image classification using meta-heuristics and artificial neural network","authors":"Sakshi Dhingra, Dharminder Kumar","doi":"10.1080/02522667.2022.2133222","DOIUrl":"https://doi.org/10.1080/02522667.2022.2133222","url":null,"abstract":"Abstract Hyperspectral images usually comprise several continuous spectral bands that represent the category of similar objects or material within the captured scene. These high-dimensional data structures have a high level of correlation and possess unique information that can be used for precise image classification. The precise selection of useful features from these high dimensional band information is very important to reduce the challenge of hyper spectral image classification approaches. Nowadays, metaheuristic algorithms are immensely utilized as a promising tool for hyperspectral image classification. In the present research work, hyperspectral images are classified with the various combinations of meta-heuristic approaches and the neural network including the mostly used Cuckoo Search (CS) optimization algorithm to resolve the global optimization search problems considering the improvement needed in image classification. Further, the strength of CS is improved using the integration of the Genetic Algorithm (GA) fitness function within the CS. The feature selection is performed by the hybrid CS and GA algorithm and the optimized features are then fed to ANN for training and classification. The paper has shown a comparative analysis of various meta heuristics techniques with ANN on parameters like kappa coefficient, Class accuracy and overall Accuracy and the designed algorithms are tested on the Indian Pines dataset. The proposed CS and GA with ANN outperformed the two already existing works with an overall average accuracy of 97.30% and a kappa coefficient of 0.9760.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2167 - 2179"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48383690","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 : 2022-11-17DOI: 10.1080/02522667.2022.2153996
Badal Acharya, Priyadarsan Parida, R. N. Panda, P. K. Mohapatra
Abstract Providing communication between two remote points via a medium that is disturbed or distorted by noise or dispersion is the purpose of a communication system. In comparison to traditional approaches, metaheuristics inspired by nature have shown better performance. In this works, Butterfly Optimization Algorithm (BOA), an algorithm inspired by nature is presented as training algorithm for ANN. Here, we apply the training strategy for BOA in channel equalization. The proposed equalizer was found to perform better than previously known NN-based equalizers based on Bit Error Rate (BER) and Mean Square Error (MSE).
{"title":"A novel approach for BOA trained ANN for channel equalization problems","authors":"Badal Acharya, Priyadarsan Parida, R. N. Panda, P. K. Mohapatra","doi":"10.1080/02522667.2022.2153996","DOIUrl":"https://doi.org/10.1080/02522667.2022.2153996","url":null,"abstract":"Abstract Providing communication between two remote points via a medium that is disturbed or distorted by noise or dispersion is the purpose of a communication system. In comparison to traditional approaches, metaheuristics inspired by nature have shown better performance. In this works, Butterfly Optimization Algorithm (BOA), an algorithm inspired by nature is presented as training algorithm for ANN. Here, we apply the training strategy for BOA in channel equalization. The proposed equalizer was found to perform better than previously known NN-based equalizers based on Bit Error Rate (BER) and Mean Square Error (MSE).","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2121 - 2130"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42538719","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 : 2022-11-17DOI: 10.1080/02522667.2022.2122199
Basim A. Hassan, Ali Ahmed A. Abdullah
Abstract The formula conjugate is usually the focal point in conjugate gradient techniques. In this paper, the Perry’s conjugacy condition and quadratic model are used to derive a new coefficient conjugate for the conjugate gradient technique, which is used to solve picture restoration issues. The algorithms show global convergence and have the required descent property. The new technique has showed substantial improvement in numerical testing. It has been demonstrated that the novel conjugate gradient approach outperforms the traditional FR conjugate gradient method. The new technique has showed substantial improvement in numerical testing. It has been demonstrated that the novel conjugate gradient approach outperforms the traditional FR conjugate gradient method.
{"title":"Enriched formulas to conjugate gradient method for removing impulse noise images","authors":"Basim A. Hassan, Ali Ahmed A. Abdullah","doi":"10.1080/02522667.2022.2122199","DOIUrl":"https://doi.org/10.1080/02522667.2022.2122199","url":null,"abstract":"Abstract The formula conjugate is usually the focal point in conjugate gradient techniques. In this paper, the Perry’s conjugacy condition and quadratic model are used to derive a new coefficient conjugate for the conjugate gradient technique, which is used to solve picture restoration issues. The algorithms show global convergence and have the required descent property. The new technique has showed substantial improvement in numerical testing. It has been demonstrated that the novel conjugate gradient approach outperforms the traditional FR conjugate gradient method. The new technique has showed substantial improvement in numerical testing. It has been demonstrated that the novel conjugate gradient approach outperforms the traditional FR conjugate gradient method.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2065 - 2074"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44942625","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 : 2022-11-17DOI: 10.1080/02522667.2022.2060910
Farah A. Abdulghani, N. A. Abdullah
Abstract Text categorization refers to the process of grouping text or documents into classes or categories according to their content, which is a significant task in natural language processing. The majority of the present work focused on English text, with a few experiments on Arabic text. The text classification process consists of many steps, from preprocessing documents (removing stop words and stem method), to feature extraction and classification phase. A new improved approach for Arabic text categorization was proposed using mutual information in a hybrid deep learning model for classification. To test the proposed model, two datasets of Arabic documents are employed. The experimental results demonstrate that employing the proposed mutual information exceeds other prior techniques in terms of performance. In Akhbarona corpus, the Multi-Layer Perceptron achieved a minimum accuracy of 96.09%, while the hybrid Convolution-Long Short-Term Memory had a performance level of 99.28%. In Khaleej corpus, the Gated Recurrent Unit had the maximum accuracy of 98.23%, while Multi-Layer Perceptron had the lowest accuracy of 97.23%
{"title":"Hybrid deep learning model for Arabic text classification based on mutual information","authors":"Farah A. Abdulghani, N. A. Abdullah","doi":"10.1080/02522667.2022.2060910","DOIUrl":"https://doi.org/10.1080/02522667.2022.2060910","url":null,"abstract":"Abstract Text categorization refers to the process of grouping text or documents into classes or categories according to their content, which is a significant task in natural language processing. The majority of the present work focused on English text, with a few experiments on Arabic text. The text classification process consists of many steps, from preprocessing documents (removing stop words and stem method), to feature extraction and classification phase. A new improved approach for Arabic text categorization was proposed using mutual information in a hybrid deep learning model for classification. To test the proposed model, two datasets of Arabic documents are employed. The experimental results demonstrate that employing the proposed mutual information exceeds other prior techniques in terms of performance. In Akhbarona corpus, the Multi-Layer Perceptron achieved a minimum accuracy of 96.09%, while the hybrid Convolution-Long Short-Term Memory had a performance level of 99.28%. In Khaleej corpus, the Gated Recurrent Unit had the maximum accuracy of 98.23%, while Multi-Layer Perceptron had the lowest accuracy of 97.23%","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"1901 - 1908"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42843974","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 : 2022-11-17DOI: 10.1080/02522667.2022.2133221
V. Kapoor, S. Dey
Abstract Recent studies in finance argue that technical analysis has the ability to predict stock prices. Though a variety of systems are used for market assessment and timing, past research has shown very little interest in optimizing the parameters of these systems. Genetic Algorithms (GA) are a soft computing based optimization procedure that optimizes a rule or parameters of a rule where search space is very large and it is not practically possible to test each and every parameter combination due to limited processing power and time. In this research we have used a GA based approach to optimize parameters of a pre-defined rule set that predicts the next-day’s stock price. Results obtained from our experiments are promising and encouraging enough to lead us to believe that Genetic Algorithm (GA) is an appropriate way of addressing these types of NP hard problems.
{"title":"A genetic algorithm based decision support system for forecasting security prices in stock index","authors":"V. Kapoor, S. Dey","doi":"10.1080/02522667.2022.2133221","DOIUrl":"https://doi.org/10.1080/02522667.2022.2133221","url":null,"abstract":"Abstract Recent studies in finance argue that technical analysis has the ability to predict stock prices. Though a variety of systems are used for market assessment and timing, past research has shown very little interest in optimizing the parameters of these systems. Genetic Algorithms (GA) are a soft computing based optimization procedure that optimizes a rule or parameters of a rule where search space is very large and it is not practically possible to test each and every parameter combination due to limited processing power and time. In this research we have used a GA based approach to optimize parameters of a pre-defined rule set that predicts the next-day’s stock price. Results obtained from our experiments are promising and encouraging enough to lead us to believe that Genetic Algorithm (GA) is an appropriate way of addressing these types of NP hard problems.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2153 - 2166"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48738507","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 : 2022-11-17DOI: 10.1080/02522667.2022.2133217
R. Chaturvedi, Udayan Ghose
Abstract In recent years object detection system has been improved by many folds due to many novel deep learning models. Deep learning has outperformed the existing traditional computer vision techniques. In recent many deep learning models uses the concept of anchor box, the model proposes various anchor boxes on the images. The models generally use a classification model and a regression models, the regression model is used to predict the position of next possible anchor box and the classification is used to validate the anchor box. The hyper tuning of these models are generally based on the anchor box specifications, many researchers have used an optimized anchor box dimensions which is obtained for a specific dataset, due to which the accuracy increases drastically but the model are not scalable on any other data set. We propose a new hybrid anchor box optimization technique by using a variant of Bayesian optimization and sub sampling for small object detection using retina net model with resnet backbone. Our hybrid model is scalable over various datasets, the model is used on visdrone dataset and the result shows a 3.7% improvement in MAP result.
{"title":"Small object detection using retinanet with hybrid anchor box hyper tuning using interface of Bayesian mathematics","authors":"R. Chaturvedi, Udayan Ghose","doi":"10.1080/02522667.2022.2133217","DOIUrl":"https://doi.org/10.1080/02522667.2022.2133217","url":null,"abstract":"Abstract In recent years object detection system has been improved by many folds due to many novel deep learning models. Deep learning has outperformed the existing traditional computer vision techniques. In recent many deep learning models uses the concept of anchor box, the model proposes various anchor boxes on the images. The models generally use a classification model and a regression models, the regression model is used to predict the position of next possible anchor box and the classification is used to validate the anchor box. The hyper tuning of these models are generally based on the anchor box specifications, many researchers have used an optimized anchor box dimensions which is obtained for a specific dataset, due to which the accuracy increases drastically but the model are not scalable on any other data set. We propose a new hybrid anchor box optimization technique by using a variant of Bayesian optimization and sub sampling for small object detection using retina net model with resnet backbone. Our hybrid model is scalable over various datasets, the model is used on visdrone dataset and the result shows a 3.7% improvement in MAP result.","PeriodicalId":46518,"journal":{"name":"JOURNAL OF INFORMATION & OPTIMIZATION SCIENCES","volume":"43 1","pages":"2099 - 2110"},"PeriodicalIF":1.4,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46051248","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：