A great amount of fake news are propagated in online social media, with the aim, usually, to deceive users and formulate specific opinions. The threat is even greater when the purpose is political or ideological and they are used during electoral campaigns. Bots play a key role in disseminating these false claims. False information is intentionally written to trigger emotions to the readers in an attempt to be believed and be disseminated in social media. Therefore, in order to discriminate credible from non credible information, we believe that it is important to take into account these emotional signals. In this paper we describe the way that emotional features have been integrated in deep learning models in order to detect if and when emotions are evoked in fake news.
{"title":"On the Impact of Emotions on the Detection of False Information","authors":"Paolo Rosso, Bilal Ghanem, Anastasia Giahanou","doi":"10.1145/3459104.3459150","DOIUrl":"https://doi.org/10.1145/3459104.3459150","url":null,"abstract":"A great amount of fake news are propagated in online social media, with the aim, usually, to deceive users and formulate specific opinions. The threat is even greater when the purpose is political or ideological and they are used during electoral campaigns. Bots play a key role in disseminating these false claims. False information is intentionally written to trigger emotions to the readers in an attempt to be believed and be disseminated in social media. Therefore, in order to discriminate credible from non credible information, we believe that it is important to take into account these emotional signals. In this paper we describe the way that emotional features have been integrated in deep learning models in order to detect if and when emotions are evoked in fake news.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131689182","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}
In order to improve the accuracy of the recommendation system, we study the influence of the number of co-rating items between users and the sequence associations between similar users on user ratings. When calculating the user similarity, we can not only consider the influence of user ratings, but also consider the similarity between user association sequences consisting of the number of co-rating items. And on this basis, we propose a more accurate user similarity measurement method, and get a more accurate user rating prediction method.The experimental results show that the proposed similarity calculation method combined with the co-rating sequence can more accurately characterize the user similarity, the user ratings prediction mean square error is smaller and the recommended effect is effectively improved compared with other algorithms. While the algorithm is based on a large number of experimental foundations, does not include deep learning into the category, so the choice of fusion coefficient may not be optimal.
{"title":"Collaborative Filtering Recommendation Algorithm Based on Similarity of Co-Rating Sequence","authors":"Xiaoyu Liu, Shuqing Li","doi":"10.1145/3459104.3459180","DOIUrl":"https://doi.org/10.1145/3459104.3459180","url":null,"abstract":"In order to improve the accuracy of the recommendation system, we study the influence of the number of co-rating items between users and the sequence associations between similar users on user ratings. When calculating the user similarity, we can not only consider the influence of user ratings, but also consider the similarity between user association sequences consisting of the number of co-rating items. And on this basis, we propose a more accurate user similarity measurement method, and get a more accurate user rating prediction method.The experimental results show that the proposed similarity calculation method combined with the co-rating sequence can more accurately characterize the user similarity, the user ratings prediction mean square error is smaller and the recommended effect is effectively improved compared with other algorithms. While the algorithm is based on a large number of experimental foundations, does not include deep learning into the category, so the choice of fusion coefficient may not be optimal.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131276592","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}
Tianqian Chen, Shuyu Chen, Shan Mei, Shuqi An, Xiaohan Yuan, Yuwen Lu
The multistep prediction of new Corona Virus Disease (COVID-19) cases plays a vital role during the epidemic control period, and the Long Short-Term Memory (LSTM) based time series analysis model is the most frequently used among many prediction methods. But whether it is the cumulative error of the multistep prediction or the instability of the new case data of the COVID-19 make the performance of LSTM in this task not so good. In this paper, we selected three countries with more severe COVID-19 epidemics—India, Russia, and Chile, to predict new cases in the next 15 days with different multistep LSTM network models, and use Bayesian Optimization to explore the optimal hyperparameter space. The results show that: a) the performance of Recursive Prediction LSTM is the best (Mean Absolute Percentage Error, MAPE was reduced to 14.88%, 6.46%, and 16.31% for the three countries respectively), Encoder Decoder LSTM is second (15.52%, 19.61%, 19.87%), and the effect of vector output LSTM is the worst (23.55%, 26.82%, 19.57%); b) there are obvious extremely poor areas in the hyperparameter space, and the Bayesian Optimizer can focus on the good areas to avoid cost of tuning parameters based on bad hyperparameters; c) the data of new cases of COVID-19 in different countries have great differences in the hyperparameter expectations for the model. The bad area of hyperparameters and different expectations are likely to be one of the reasons why the COVID-19 data of different countries is hard to train jointly.
{"title":"Multistep Forecasting of New COVID-19 Cases Based on LSTMs Using Bayesian Optimization","authors":"Tianqian Chen, Shuyu Chen, Shan Mei, Shuqi An, Xiaohan Yuan, Yuwen Lu","doi":"10.1145/3459104.3459116","DOIUrl":"https://doi.org/10.1145/3459104.3459116","url":null,"abstract":"The multistep prediction of new Corona Virus Disease (COVID-19) cases plays a vital role during the epidemic control period, and the Long Short-Term Memory (LSTM) based time series analysis model is the most frequently used among many prediction methods. But whether it is the cumulative error of the multistep prediction or the instability of the new case data of the COVID-19 make the performance of LSTM in this task not so good. In this paper, we selected three countries with more severe COVID-19 epidemics—India, Russia, and Chile, to predict new cases in the next 15 days with different multistep LSTM network models, and use Bayesian Optimization to explore the optimal hyperparameter space. The results show that: a) the performance of Recursive Prediction LSTM is the best (Mean Absolute Percentage Error, MAPE was reduced to 14.88%, 6.46%, and 16.31% for the three countries respectively), Encoder Decoder LSTM is second (15.52%, 19.61%, 19.87%), and the effect of vector output LSTM is the worst (23.55%, 26.82%, 19.57%); b) there are obvious extremely poor areas in the hyperparameter space, and the Bayesian Optimizer can focus on the good areas to avoid cost of tuning parameters based on bad hyperparameters; c) the data of new cases of COVID-19 in different countries have great differences in the hyperparameter expectations for the model. The bad area of hyperparameters and different expectations are likely to be one of the reasons why the COVID-19 data of different countries is hard to train jointly.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133849853","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}
P. Karmalkar, H. Gurulingappa, Justna Muhith, Shikha Singhal, Gerard Megaro, F. Buchholz
Detecting language nuances from unstructured data could be the difference in serving up the right Google search results or using unsolicited social media chatter to tap into unexplored customer behavior (patients and HCPs). However, as an established science, there is a slow adoption of NLP and Text Analytics in healthcare sector for analysis of unstructured textual data originating from customer interactions. One of the key areas of our exploration is the Medical Information function within our organization. We receive a significant amount of medical information inquires in the form of unstructured data through multiple communication channels. The current system of gathering insights takes significant time and effort – as information must be manually tagged and classified limiting the ability to drive insights and trends efficiently and in a timely manner. These limitations mean subject matter experts must spend time manually deducing insights and aligning with medical affairs – time that could be better spent elsewhere. Therefore, this article presents an approach using NLP & Text Analytics to generate valuable insights from unstructured medical information inquiries. The system automatically extracts key phrases, medical terms, themes, sentiments as well as leverages unsupervised statistical modeling for two-level categorization of inquiries. Results of NLP when analyzed with the aid of visual analytics tool highlighted non-obvious insights indicating the value it can generate to influence product strategies.
{"title":"Improving Consumer Experience for Medical Information Using Text Analytics","authors":"P. Karmalkar, H. Gurulingappa, Justna Muhith, Shikha Singhal, Gerard Megaro, F. Buchholz","doi":"10.1145/3459104.3459182","DOIUrl":"https://doi.org/10.1145/3459104.3459182","url":null,"abstract":"Detecting language nuances from unstructured data could be the difference in serving up the right Google search results or using unsolicited social media chatter to tap into unexplored customer behavior (patients and HCPs). However, as an established science, there is a slow adoption of NLP and Text Analytics in healthcare sector for analysis of unstructured textual data originating from customer interactions. One of the key areas of our exploration is the Medical Information function within our organization. We receive a significant amount of medical information inquires in the form of unstructured data through multiple communication channels. The current system of gathering insights takes significant time and effort – as information must be manually tagged and classified limiting the ability to drive insights and trends efficiently and in a timely manner. These limitations mean subject matter experts must spend time manually deducing insights and aligning with medical affairs – time that could be better spent elsewhere. Therefore, this article presents an approach using NLP & Text Analytics to generate valuable insights from unstructured medical information inquiries. The system automatically extracts key phrases, medical terms, themes, sentiments as well as leverages unsupervised statistical modeling for two-level categorization of inquiries. Results of NLP when analyzed with the aid of visual analytics tool highlighted non-obvious insights indicating the value it can generate to influence product strategies.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115195370","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}
Over recent years, tensors have emerged as the preferred data structure for model representation and computation in machine learning. However, current tensor models suffer from a lack of a formal basis, where the tensors are treated as arbitrary multidimensional data processed by a large and ever-growing collection of functions added ad hoc. In this way, tensor frameworks degenerate to programming languages with a curiously cumbersome data model. This paper argues that a more formal basis for tensors and their computation brings important benefits. The proposed formalism is based on 1) a strong type system for tensors with named dimensions, 2) a common model of both dense and sparse tensors, and 3) a small, closed set of tensor functions, providing a general mathematical language in which higher level functions can be expressed. These features work together to provide ease of use resulting from static type verification with meaningful dimension names, improved interoperability resulting from defining a closed set of just six foundational tensor functions, and better support for performance optimizations resulting from having just a small set of core functions needing low-level optimizations, and higher-level operations being able to work on arbitrary chunks of these functions, as well as from better mathematical properties from using named tensor dimensions without inherent order. The proposed model is implemented as the model inference engine in the Vespa big data serving engine, where it runs various models expressed in this language directly, as well as models expressed in TensorFlow or Onnx formats.
{"title":"A Tensor Formalism for Computer Science","authors":"Jon Bratseth, H. Pettersen, L. Solbakken","doi":"10.1145/3459104.3459152","DOIUrl":"https://doi.org/10.1145/3459104.3459152","url":null,"abstract":"Over recent years, tensors have emerged as the preferred data structure for model representation and computation in machine learning. However, current tensor models suffer from a lack of a formal basis, where the tensors are treated as arbitrary multidimensional data processed by a large and ever-growing collection of functions added ad hoc. In this way, tensor frameworks degenerate to programming languages with a curiously cumbersome data model. This paper argues that a more formal basis for tensors and their computation brings important benefits. The proposed formalism is based on 1) a strong type system for tensors with named dimensions, 2) a common model of both dense and sparse tensors, and 3) a small, closed set of tensor functions, providing a general mathematical language in which higher level functions can be expressed. These features work together to provide ease of use resulting from static type verification with meaningful dimension names, improved interoperability resulting from defining a closed set of just six foundational tensor functions, and better support for performance optimizations resulting from having just a small set of core functions needing low-level optimizations, and higher-level operations being able to work on arbitrary chunks of these functions, as well as from better mathematical properties from using named tensor dimensions without inherent order. The proposed model is implemented as the model inference engine in the Vespa big data serving engine, where it runs various models expressed in this language directly, as well as models expressed in TensorFlow or Onnx formats.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124004183","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}
In order to analyze the influence of noise environment on the structural performance of airborne equipment, this paper presents an investigation on the application of structural noise bearing capability analysis method in airborne radar equipment. In this investigation, a method for analyzing the structural noise bearing capability is introduced, which could convert the acoustic excitation load to the structural random vibration load. Based on this method, the transformation of noise spectrum into the sound pressure power spectrum density is realized. Additionally, the structural performance of airborne radar equipment under noise environment is analyzed by finite element simulation, and the feasibility of applying the structural noise bearing capability analysis method in airborne radar equipment is verified. This study has certain significance in engineering practice for the analysis of noise environment adaptability of equipment.
{"title":"Investigation on the Application of Structural Noise Bearing Capability Analysis Method in Airborne Radar Equipment","authors":"Z. Gu, Zhigang Qin","doi":"10.1145/3459104.3459107","DOIUrl":"https://doi.org/10.1145/3459104.3459107","url":null,"abstract":"In order to analyze the influence of noise environment on the structural performance of airborne equipment, this paper presents an investigation on the application of structural noise bearing capability analysis method in airborne radar equipment. In this investigation, a method for analyzing the structural noise bearing capability is introduced, which could convert the acoustic excitation load to the structural random vibration load. Based on this method, the transformation of noise spectrum into the sound pressure power spectrum density is realized. Additionally, the structural performance of airborne radar equipment under noise environment is analyzed by finite element simulation, and the feasibility of applying the structural noise bearing capability analysis method in airborne radar equipment is verified. This study has certain significance in engineering practice for the analysis of noise environment adaptability of equipment.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"2016 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127374085","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}
Securing the traceability of products in a supply chain is an urgent issue. Recently, supply-chain systems that use a blockchain have been proposed. In these systems, the blockchain is used as a common database shared among supply chain parties to secure the integrity and reliability of distribution information such as ownership transfer records. These systems thus secure a high level of traceability in the supply chain. Considering future scalability of supply chains, public permissionless blockchain (PPBC) is a promising approach. In this approach, however, distribution information that should be kept private is made public since the information recorded in PPBC can be read by anyone. We therefore propose a method for preserving privacy while securing traceability in a supply chain system using PPBC. The proposed method preserves privacy by concealing distribution information via encryption. In addition, the proposed method ensures distribution among legitimate supply chain parties while concealing their blockchain addresses by using zero-knowledge proofs.We implement the proposed method on Ethereum smart contracts and verify the system behavior. The results show that the proposed method works as expected, and that system usage cost per distribution party is at most 2.2 × 106 gas units in terms of blockchain transaction fees.
{"title":"Design and Evaluation of a Privacy-preserving Supply Chain System Based on Public Permissionless Blockchain","authors":"Takio Uesugi, Yoshinobu Shijo, M. Murata","doi":"10.1145/3459104.3459155","DOIUrl":"https://doi.org/10.1145/3459104.3459155","url":null,"abstract":"Securing the traceability of products in a supply chain is an urgent issue. Recently, supply-chain systems that use a blockchain have been proposed. In these systems, the blockchain is used as a common database shared among supply chain parties to secure the integrity and reliability of distribution information such as ownership transfer records. These systems thus secure a high level of traceability in the supply chain. Considering future scalability of supply chains, public permissionless blockchain (PPBC) is a promising approach. In this approach, however, distribution information that should be kept private is made public since the information recorded in PPBC can be read by anyone. We therefore propose a method for preserving privacy while securing traceability in a supply chain system using PPBC. The proposed method preserves privacy by concealing distribution information via encryption. In addition, the proposed method ensures distribution among legitimate supply chain parties while concealing their blockchain addresses by using zero-knowledge proofs.We implement the proposed method on Ethereum smart contracts and verify the system behavior. The results show that the proposed method works as expected, and that system usage cost per distribution party is at most 2.2 × 106 gas units in terms of blockchain transaction fees.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128882211","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 employs multiscale feature extraction based on Simpson's diversity index for predicting remaining useful life (RUL) of bearings. Being a measure of variety of elements in the given time series, Simpson's diversity index (SDI) acts as a feature which is assumed to be different for time series of different quality. Thus, RUL is considered to be function of multiscale SDI in this paper. Features are mapped to RUL with modified Tensor product Bernstein polynomial (TPBP) network. The aim of this paper is to test SDI based feature extraction together with modified TPBP network for in the context of RUL analysis.
{"title":"Multiscale Diversity Index for RUL Analysis with Bernstein Polynomial Neural Networks","authors":"M. Landauskas, L. Saunoriene, M. Ragulskis","doi":"10.1145/3459104.3459188","DOIUrl":"https://doi.org/10.1145/3459104.3459188","url":null,"abstract":"This paper employs multiscale feature extraction based on Simpson's diversity index for predicting remaining useful life (RUL) of bearings. Being a measure of variety of elements in the given time series, Simpson's diversity index (SDI) acts as a feature which is assumed to be different for time series of different quality. Thus, RUL is considered to be function of multiscale SDI in this paper. Features are mapped to RUL with modified Tensor product Bernstein polynomial (TPBP) network. The aim of this paper is to test SDI based feature extraction together with modified TPBP network for in the context of RUL analysis.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133690700","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The purpose of this study is to work efficiently with the customer relations management "Big Data" of corporations on a global scale. We asked 15 companies about 15 questions about customer relationship management. As a result of 10 article reviews, we examined customer relationship management data. Descriptive research design was used in the research with a quantitative research approach. The collected data were analyzed with inferential statistics using linear regression as descriptive statistics. When the researches on "customer relationship management" are examined, the complaints of the customers and the problems of customer loss are discussed. There are suggestions for the problems examined. However, this study describes a model that will support these recommendations and take them forward. From the research conducted so far, it is understood that there are difficulties in the use of technology. The association, which is planned as the Customer Relationship Management Association, will serve as a framework for the customer relations units of institutions using "Big Data". Customer acquisition and development is targeted by combining organizations with technology and manpower in customer relationship management.
{"title":"Customer Relationship Management Association Establishing A Customer Relationship Management Association That Will Act As A Roof With Big Data","authors":"Ceren Ulak","doi":"10.1145/3459104.3459121","DOIUrl":"https://doi.org/10.1145/3459104.3459121","url":null,"abstract":"The purpose of this study is to work efficiently with the customer relations management \"Big Data\" of corporations on a global scale. We asked 15 companies about 15 questions about customer relationship management. As a result of 10 article reviews, we examined customer relationship management data. Descriptive research design was used in the research with a quantitative research approach. The collected data were analyzed with inferential statistics using linear regression as descriptive statistics. When the researches on \"customer relationship management\" are examined, the complaints of the customers and the problems of customer loss are discussed. There are suggestions for the problems examined. However, this study describes a model that will support these recommendations and take them forward. From the research conducted so far, it is understood that there are difficulties in the use of technology. The association, which is planned as the Customer Relationship Management Association, will serve as a framework for the customer relations units of institutions using \"Big Data\". Customer acquisition and development is targeted by combining organizations with technology and manpower in customer relationship management.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"130 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127382992","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}
Recognizing faces is a very challenging problem in the field of image processing. Deep neural network and especially Convolutional Neural Networks are the most widely used techniques for image classification and recognition. Despite these deep neural networks efficiency, choosing their optimal architectures for a given task remains an open problem. In fact, Convolutional Neural Networks performance depends on many hyper-parameters namely the network depth, convolutional layer numbers, the number of the local receptive fields and their respective sizes, convolutional stride and dropout ratio. These parameters thoroughly affect the performance of the classifier. This paper aims to optimize these parameters and develop the optimized architecture face classification and recognition. Intensive simulated experiments and qualitative comparisons have been conducted. The achieved results show that the developed Convolutional Neural Networks configuration provided a remarkable performance improvement in in terms of the network accuracy that exceeds 94%.
{"title":"Deep Neural Network Hyper-Parameters Optimization for Face Classification","authors":"M. Awadalla, A. Galal","doi":"10.1145/3459104.3459143","DOIUrl":"https://doi.org/10.1145/3459104.3459143","url":null,"abstract":"Recognizing faces is a very challenging problem in the field of image processing. Deep neural network and especially Convolutional Neural Networks are the most widely used techniques for image classification and recognition. Despite these deep neural networks efficiency, choosing their optimal architectures for a given task remains an open problem. In fact, Convolutional Neural Networks performance depends on many hyper-parameters namely the network depth, convolutional layer numbers, the number of the local receptive fields and their respective sizes, convolutional stride and dropout ratio. These parameters thoroughly affect the performance of the classifier. This paper aims to optimize these parameters and develop the optimized architecture face classification and recognition. Intensive simulated experiments and qualitative comparisons have been conducted. The achieved results show that the developed Convolutional Neural Networks configuration provided a remarkable performance improvement in in terms of the network accuracy that exceeds 94%.","PeriodicalId":142284,"journal":{"name":"2021 International Symposium on Electrical, Electronics and Information Engineering","volume":"982 ","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120942404","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：