Pub Date : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375240
Sunsern Ceamanunkul, Sanchit Chawla
Drowsy drivers are a major cause of many road accidents around the world. Facial emotions are known to be one of the visual cues for detecting drowsiness. In this paper, we propose a machine learning approach to drowsiness detection based on using a combination of facial emotion features extracted by using deep convolutional neural networks (CNN) and eye-aspect-ratio (EAR) features. The combined feature vectors are then used for training a classifier. From our experiments, we obtain a classification accuracy of 81.7% when we use the combined features with a support vector machines (SVM) classifier.
{"title":"Drowsiness Detection using Facial Emotions and Eye Aspect Ratios","authors":"Sunsern Ceamanunkul, Sanchit Chawla","doi":"10.1109/ICSEC51790.2020.9375240","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375240","url":null,"abstract":"Drowsy drivers are a major cause of many road accidents around the world. Facial emotions are known to be one of the visual cues for detecting drowsiness. In this paper, we propose a machine learning approach to drowsiness detection based on using a combination of facial emotion features extracted by using deep convolutional neural networks (CNN) and eye-aspect-ratio (EAR) features. The combined feature vectors are then used for training a classifier. From our experiments, we obtain a classification accuracy of 81.7% when we use the combined features with a support vector machines (SVM) classifier.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122071346","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375423
Kasi Tenghongsakul, Isoon Kanjanasurat, B. Purahong, A. Lasakul
At present, many of visual disease happened from the abnormality of retinal vessels. The automatic vascular extraction from fundus images is essential for the diagnosis to reduce vision loss. This paper offers retinal blood vessel segmentation using the pre-processing and IterNet model, a convolution neural network. The green channel and gray scale image that is high contrast between the blood vessel and background, including the normalization, were used to improve blood vessel image quality. The proposed method was tested with two widely used databases, including DRIVE and CHASEDB-1, which unique characteristics in each data set. The results of blood vessel extraction of Drive and CHASEDB-1 achieved sensitivity 0.8126 and 0.7541, respectively.
{"title":"Retinal Blood Vessel Extraction by Using Pre-processing and IterNet Model","authors":"Kasi Tenghongsakul, Isoon Kanjanasurat, B. Purahong, A. Lasakul","doi":"10.1109/ICSEC51790.2020.9375423","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375423","url":null,"abstract":"At present, many of visual disease happened from the abnormality of retinal vessels. The automatic vascular extraction from fundus images is essential for the diagnosis to reduce vision loss. This paper offers retinal blood vessel segmentation using the pre-processing and IterNet model, a convolution neural network. The green channel and gray scale image that is high contrast between the blood vessel and background, including the normalization, were used to improve blood vessel image quality. The proposed method was tested with two widely used databases, including DRIVE and CHASEDB-1, which unique characteristics in each data set. The results of blood vessel extraction of Drive and CHASEDB-1 achieved sensitivity 0.8126 and 0.7541, respectively.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128818333","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375275
Kasemsit Teeyapan
Abnormality detection in musculoskeletal radiographs, a regular task for radiologists, requires both experiences and efforts. To increase the number of radiographs interpreted each day, this paper presents cost-efficient deep learning models based on ensembles of EfficientNet architectures to help automate the detection process. We investigate the transfer learning performance of ImageNet pre-trained checkpoints on the musculoskeletal radiograph (MURA) dataset which is very different from the ImageNet dataset. The experimental results show that, the ImageNet pre-trained checkpoints have to be retrained on the entire MURA training set, before being trained on a specific study type. The performance of the EfficientNet-based models is shown to be superior to three baseline models. In particular, EfficientNet-B3 not only achieved the overall Cohen’s Kappa score of 0.717, compared to the scores 0.680, 0.688, and 0.712 for MobileNetV2, DenseNet-169, and Xception, respectively, but also being better in term of efficiency.
{"title":"Abnormality Detection in Musculoskeletal Radiographs using EfficientNets","authors":"Kasemsit Teeyapan","doi":"10.1109/ICSEC51790.2020.9375275","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375275","url":null,"abstract":"Abnormality detection in musculoskeletal radiographs, a regular task for radiologists, requires both experiences and efforts. To increase the number of radiographs interpreted each day, this paper presents cost-efficient deep learning models based on ensembles of EfficientNet architectures to help automate the detection process. We investigate the transfer learning performance of ImageNet pre-trained checkpoints on the musculoskeletal radiograph (MURA) dataset which is very different from the ImageNet dataset. The experimental results show that, the ImageNet pre-trained checkpoints have to be retrained on the entire MURA training set, before being trained on a specific study type. The performance of the EfficientNet-based models is shown to be superior to three baseline models. In particular, EfficientNet-B3 not only achieved the overall Cohen’s Kappa score of 0.717, compared to the scores 0.680, 0.688, and 0.712 for MobileNetV2, DenseNet-169, and Xception, respectively, but also being better in term of efficiency.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"138 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122055536","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}
Agricultural research is a very important activity for developing countries as its economy relies on the agricultural sector. To ensure that the investment in the research is in the right direction, it is necessary to determine the relationship between trade values and invested research. However, the effective and efficient evaluation is constrained by the complexity and fragmentation of information required for analysis. The large number of agricultural products and related research items occurred between the time research grants were allocated and the time of the trade, such as research projects, publications, intellectual property, etc. mean that the amount of data to be processed is enormous and is responsible by many organizations. The data which are collected and stored in different databases are uncoordinated and there are seldom explicit links between records, both within and across databases. The only research item with direct links is research publication and even that is rarely attributed directly to research grants.In this paper, we propose a framework for cross-datasources analysis for agricultural products. The data are automatically collected from official sources of agricultural data and stored into a unified database to eliminate dependencies between the visualization and structure of datasources. The pathways are recognized by analyzing links between items among their parameters, such as names, affiliations, etc. The framework is demonstrated by analyzing agricultural research activities in Thailand. The total number of gathered data records is approximately 8.8 million records. Visualization of research-to-impact pathways of two agricultural products (pineapple and sugarcane) are used as case study.
{"title":"A framework for cross-datasources agricultural research-to-impact analysis","authors":"Nalina Phisanbut, Poonsak Nuchsiri, Pasith Thanapatpisarn, Sittinun Pinthaya, Noppagorn Panpa, Piyanat Teinlek, P. Piamsa-nga","doi":"10.1109/ICSEC51790.2020.9375271","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375271","url":null,"abstract":"Agricultural research is a very important activity for developing countries as its economy relies on the agricultural sector. To ensure that the investment in the research is in the right direction, it is necessary to determine the relationship between trade values and invested research. However, the effective and efficient evaluation is constrained by the complexity and fragmentation of information required for analysis. The large number of agricultural products and related research items occurred between the time research grants were allocated and the time of the trade, such as research projects, publications, intellectual property, etc. mean that the amount of data to be processed is enormous and is responsible by many organizations. The data which are collected and stored in different databases are uncoordinated and there are seldom explicit links between records, both within and across databases. The only research item with direct links is research publication and even that is rarely attributed directly to research grants.In this paper, we propose a framework for cross-datasources analysis for agricultural products. The data are automatically collected from official sources of agricultural data and stored into a unified database to eliminate dependencies between the visualization and structure of datasources. The pathways are recognized by analyzing links between items among their parameters, such as names, affiliations, etc. The framework is demonstrated by analyzing agricultural research activities in Thailand. The total number of gathered data records is approximately 8.8 million records. Visualization of research-to-impact pathways of two agricultural products (pineapple and sugarcane) are used as case study.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121197445","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375214
Mozammel H. A. Khan
Data clustering algorithms partition a given set of data points into groups containing very similar data points. Representative-based and density-based algorithms are generally used for data clustering. These algorithms are heuristic algorithms and may stuck at a sub-optimal clustering. Crisp clustering problem is a combinatorial optimization problem. Genetic Algorithms generally perform better than heuristic algorithms for combinatorial optimization. In this work, we propose a hybrid Genetic Algorithm for density-based clustering. For this purpose, we represent a cluster using a forest of trees, where the nodes of the trees are the data points. We use a tree-based fitness function. Beside 1-point crossover, we use a deterministic improvement of offspring. We implement the proposed algorithm using C language and run on a personal computer. We experiment with five datasets from UCI Machine Learning Repository. The proposed algorithm outperforms for both low and high-dimensional datasets over existing algorithms, except for one high-dimensional dataset.
{"title":"Tree-Based Hybrid Genetic Algorithm for Density-Based Data Clustering","authors":"Mozammel H. A. Khan","doi":"10.1109/ICSEC51790.2020.9375214","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375214","url":null,"abstract":"Data clustering algorithms partition a given set of data points into groups containing very similar data points. Representative-based and density-based algorithms are generally used for data clustering. These algorithms are heuristic algorithms and may stuck at a sub-optimal clustering. Crisp clustering problem is a combinatorial optimization problem. Genetic Algorithms generally perform better than heuristic algorithms for combinatorial optimization. In this work, we propose a hybrid Genetic Algorithm for density-based clustering. For this purpose, we represent a cluster using a forest of trees, where the nodes of the trees are the data points. We use a tree-based fitness function. Beside 1-point crossover, we use a deterministic improvement of offspring. We implement the proposed algorithm using C language and run on a personal computer. We experiment with five datasets from UCI Machine Learning Repository. The proposed algorithm outperforms for both low and high-dimensional datasets over existing algorithms, except for one high-dimensional dataset.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129820468","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375279
Tapanapong Chuntama, P. Techa-angkoon, C. Suwannajak, Benjamas Panyangam, N. Tanakul
Data in astronomy usually contain various classes of astronomical objects. In this study, we explore the application of multiclass classification in classifying astronomical objects in the galaxy MS1. Our objective is to specify machine learning techniques that are best suited to our data and our classification goal. We used the archival data retrieved from the CanadaFrance-Hawaii Telescope (CFHT) data archive. The imaging data were transformed into data tables, then classified based on their visual appearances into five classes, including star, globular cluster, rounded galaxy, elongated galaxy, and fuzzy object. The classified data were used for supervised machine learning model building and testing. We investigated seven classification techniques, including Random Forest, Multilayer Perceptron, Weightless neural network (WiSARD), Deep learning (Weka deep learning), Logistic Regression, Support Vector Machine (SVM), and Multiclass Classifier. Our experiments show that Random Forest and Multilayer Perceptron archived the highest overall performances and are the best-suited model for classifying astronomical objects in the CFHT data of the galaxy M81.
{"title":"Multiclass Classification of Astronomical Objects in the Galaxy M81 using Machine Learning Techniques","authors":"Tapanapong Chuntama, P. Techa-angkoon, C. Suwannajak, Benjamas Panyangam, N. Tanakul","doi":"10.1109/ICSEC51790.2020.9375279","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375279","url":null,"abstract":"Data in astronomy usually contain various classes of astronomical objects. In this study, we explore the application of multiclass classification in classifying astronomical objects in the galaxy MS1. Our objective is to specify machine learning techniques that are best suited to our data and our classification goal. We used the archival data retrieved from the CanadaFrance-Hawaii Telescope (CFHT) data archive. The imaging data were transformed into data tables, then classified based on their visual appearances into five classes, including star, globular cluster, rounded galaxy, elongated galaxy, and fuzzy object. The classified data were used for supervised machine learning model building and testing. We investigated seven classification techniques, including Random Forest, Multilayer Perceptron, Weightless neural network (WiSARD), Deep learning (Weka deep learning), Logistic Regression, Support Vector Machine (SVM), and Multiclass Classifier. Our experiments show that Random Forest and Multilayer Perceptron archived the highest overall performances and are the best-suited model for classifying astronomical objects in the CFHT data of the galaxy M81.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"206 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116192172","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375142
Thapani Hengsanankun, Atchara Namburi
Word segmentation is a basic problem in the natural language processing of non-boundary delimiters language, especially for the Thai language. The ambiguity of the boundaries of the words in the sentence is one of the significant problems that can cause an unknown word and affects the word segmentation accuracy. This paper presents an improving Thai word segmentation using Hidden Markov Model to cope with an unknown word problem. The five-state of left-to-right HMMs are built according to the classes of the unknown word by applied the parts of speech of the Thai language as the observation symbols of the model. To determine the unknown word in the sentence, the String Matching algorithm is first implemented to find overlapping words and unknown words. The unknown words that unidentified by the lexical dictionary are classified according to their classes by the HMMs. Then the word combining rules are applied to determine the proper word boundary and to merge possible characters into words. In addition, the sentiment analysis task of polarity detection was selected as a case study to verify the accuracy of the proposed method. The precision, recall, and F-measure are used for evaluating the efficiency of the proposed method. The empirical results show that both segmented words and polarity classification results obtained by the proposed method tend to outperform the existing methods.
{"title":"Improving Thai Word Segmentation using HMM: A Case Study of Sentiment Analysis","authors":"Thapani Hengsanankun, Atchara Namburi","doi":"10.1109/ICSEC51790.2020.9375142","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375142","url":null,"abstract":"Word segmentation is a basic problem in the natural language processing of non-boundary delimiters language, especially for the Thai language. The ambiguity of the boundaries of the words in the sentence is one of the significant problems that can cause an unknown word and affects the word segmentation accuracy. This paper presents an improving Thai word segmentation using Hidden Markov Model to cope with an unknown word problem. The five-state of left-to-right HMMs are built according to the classes of the unknown word by applied the parts of speech of the Thai language as the observation symbols of the model. To determine the unknown word in the sentence, the String Matching algorithm is first implemented to find overlapping words and unknown words. The unknown words that unidentified by the lexical dictionary are classified according to their classes by the HMMs. Then the word combining rules are applied to determine the proper word boundary and to merge possible characters into words. In addition, the sentiment analysis task of polarity detection was selected as a case study to verify the accuracy of the proposed method. The precision, recall, and F-measure are used for evaluating the efficiency of the proposed method. The empirical results show that both segmented words and polarity classification results obtained by the proposed method tend to outperform the existing methods.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"235 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132424837","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375153
Pongsakorn Ajchariyasakchai, T. Rakthanmanon
Time series shapelets is a snippets of time series that can distinguish one class from others. In the last decade, many researches show that time series shapelets is not only one of the most promising classification techniques, but also a desirable solution because it is simply an explainable result to the experts. However, Two main drawbacks of time series shapelets discovery are speed and the appearance of the candidates and its representative, i.e. the time series shapelets itself. In this paper, we do not improve the running time of discovering the time series shapelets, but we propose a new method to learn the shape of time series shapelets, instead of picking one from candidates. The number of candidates can be vary from ten thousands to millions subsequences or even more depended on the length of the candidates. In this paper, autoencoder technique is applied to reduce the complexity of candidates from the higher-dimensional space to the much smaller-dimensional space, to highlight the potential candidates as the representatives, to learn the shapes of those candidates instead of the individual one, and to reconstruct the more-smooth time series shapelets. Our time series shapelets, named autoshaplets, is not fit to the exact value of the training data anymore, which normally is noisy according to the real observation. The experiment results demonstrate that the new generated shapelets can achieve higher accuracy compared to the exact shapelets, and it is less sensitive to the training data.
{"title":"AutoShapelet: Reconstructable Time Series Shapelets","authors":"Pongsakorn Ajchariyasakchai, T. Rakthanmanon","doi":"10.1109/ICSEC51790.2020.9375153","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375153","url":null,"abstract":"Time series shapelets is a snippets of time series that can distinguish one class from others. In the last decade, many researches show that time series shapelets is not only one of the most promising classification techniques, but also a desirable solution because it is simply an explainable result to the experts. However, Two main drawbacks of time series shapelets discovery are speed and the appearance of the candidates and its representative, i.e. the time series shapelets itself. In this paper, we do not improve the running time of discovering the time series shapelets, but we propose a new method to learn the shape of time series shapelets, instead of picking one from candidates. The number of candidates can be vary from ten thousands to millions subsequences or even more depended on the length of the candidates. In this paper, autoencoder technique is applied to reduce the complexity of candidates from the higher-dimensional space to the much smaller-dimensional space, to highlight the potential candidates as the representatives, to learn the shapes of those candidates instead of the individual one, and to reconstruct the more-smooth time series shapelets. Our time series shapelets, named autoshaplets, is not fit to the exact value of the training data anymore, which normally is noisy according to the real observation. The experiment results demonstrate that the new generated shapelets can achieve higher accuracy compared to the exact shapelets, and it is less sensitive to the training data.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131552318","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 : 2020-12-03DOI: 10.1109/ICSEC51790.2020.9375339
Myo Mar Thinn, Ye Kyaw Thu, Hlaing Myat Nwe, Nyo Nyo Yee, Thandar Myint, Hninn Aye Thant, T. Supnithi
This paper describes the machine translation of LATEX encoded mathematical equations to spoken mathematical sentences. A LATEX- Spoken math parallel corpus (5,600 sentences) was developed. In this paper, the 10-fold cross-validation experiments were carried out by applying Phrase-based Statistical Machine Translation (PBSMT), Weighted Finite-State Transducers (WFST) and Ripple Down Rules (RDR) based tagging approaches. The BLEU, RIBES, F1 and WER evaluation scoring metrics are used for measuring translation performance. The experimental results show that the PBSMT approach achieved the highest translation performance for LATEX mathematical equations to spoken mathematical sentences translation. Moreover, we found that the translation performance of RDR approach is comparable with PBSMT.
本文描述了LATEX编码数学方程到口语数学句子的机器翻译。开发了一个数学口语平行语料库(5600个句子)。本文采用基于短语的统计机器翻译(PBSMT)、加权有限状态传感器(WFST)和Ripple Down Rules (RDR)标记方法进行了10倍交叉验证实验。BLEU, RIBES, F1和WER评估评分指标用于衡量翻译性能。实验结果表明，PBSMT方法对LATEX数学方程到口语数学句子的翻译效果最好。此外，我们发现RDR方法的翻译性能与PBSMT方法相当。
{"title":"Machine Translation of LATEX Based Mathematical Equations to Spoken Mathematics","authors":"Myo Mar Thinn, Ye Kyaw Thu, Hlaing Myat Nwe, Nyo Nyo Yee, Thandar Myint, Hninn Aye Thant, T. Supnithi","doi":"10.1109/ICSEC51790.2020.9375339","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375339","url":null,"abstract":"This paper describes the machine translation of LATEX encoded mathematical equations to spoken mathematical sentences. A LATEX- Spoken math parallel corpus (5,600 sentences) was developed. In this paper, the 10-fold cross-validation experiments were carried out by applying Phrase-based Statistical Machine Translation (PBSMT), Weighted Finite-State Transducers (WFST) and Ripple Down Rules (RDR) based tagging approaches. The BLEU, RIBES, F1 and WER evaluation scoring metrics are used for measuring translation performance. The experimental results show that the PBSMT approach achieved the highest translation performance for LATEX mathematical equations to spoken mathematical sentences translation. Moreover, we found that the translation performance of RDR approach is comparable with PBSMT.","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130121913","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 considers dominating set problems in a disk graphs, which is a generalization of unit disk graphs extensively used to analyze homogeneous sensor or wireless networks. When considering heterogeneous networks, it is useful to consider disk graphs that contain disks with different radii. Given graph $G=(V,E)$, set $Dsubseteq V$ is a $(k,m)$ -connected dominating set for G if every node in V is either in D or has at least m neighbors in D and the induced subgraph $G[D]$ is also k -connected. Many approximation algorithms are known for this problem in unit disk graphs. We prove various properties for disk graphs so that these algorithms can be generalized to disk graphs. Namely, we show that a $displaystyle minleft{frac{m}{m-k},sqrt{k}right}cdot Oleft(ln^{2}kright)$ - approximation algorithm of Nutov works in this setting. We also present a PTAS for finding a $(1+epsilon)$ -approximate solution to the m -dominating set problem in disk graphs that runs in time $n^{O(m/epsilon)}$
{"title":"Approximating k-Connected m-Dominating Sets in Disk Graphs","authors":"Kunanon Burathep, Jittat Fakcharoenphol, Nonthaphat Wongwattanakij","doi":"10.1109/ICSEC51790.2020.9375178","DOIUrl":"https://doi.org/10.1109/ICSEC51790.2020.9375178","url":null,"abstract":"This paper considers dominating set problems in a disk graphs, which is a generalization of unit disk graphs extensively used to analyze homogeneous sensor or wireless networks. When considering heterogeneous networks, it is useful to consider disk graphs that contain disks with different radii. Given graph $G=(V,E)$, set $Dsubseteq V$ is a $(k,m)$ -connected dominating set for G if every node in V is either in D or has at least m neighbors in D and the induced subgraph $G[D]$ is also k -connected. Many approximation algorithms are known for this problem in unit disk graphs. We prove various properties for disk graphs so that these algorithms can be generalized to disk graphs. Namely, we show that a $displaystyle minleft{frac{m}{m-k},sqrt{k}right}cdot Oleft(ln^{2}kright)$ - approximation algorithm of Nutov works in this setting. We also present a PTAS for finding a $(1+epsilon)$ -approximate solution to the m -dominating set problem in disk graphs that runs in time $n^{O(m/epsilon)}$","PeriodicalId":158728,"journal":{"name":"2020 24th International Computer Science and Engineering Conference (ICSEC)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123279432","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：