Aiming at the problems of low recognition rate and slow training speed of facial expression recognition method in the background of complex images, an improved facial expression recognition algorithm based on convolutional neural networks is proposed. The proposed algorithm introduces K-Means clustering idea and SVM classifier in the framework of convolutional neural network. Firstly, the algorithm trains the K-Means clustering model by using the label-free expression images, and selects the K-means clustering centers with good data characteristics, which are used as the initial value of the convolution kernel of the CNN model to extract features. Secondly, using the feature extraction processing of the convolutional neural network, the extracted features are fed to the multi-class SVM classifier. The experimental results show that the proposed method reduces the training time of the model overall, improves the accuracies of facial expression recognition under the background of complex images, and has a certain robustness.
{"title":"Facial Expression Recognition Algorithm Based on the Combination of CNN and K-Means","authors":"Tongtong Cao, Ming Li","doi":"10.1145/3318299.3318344","DOIUrl":"https://doi.org/10.1145/3318299.3318344","url":null,"abstract":"Aiming at the problems of low recognition rate and slow training speed of facial expression recognition method in the background of complex images, an improved facial expression recognition algorithm based on convolutional neural networks is proposed. The proposed algorithm introduces K-Means clustering idea and SVM classifier in the framework of convolutional neural network. Firstly, the algorithm trains the K-Means clustering model by using the label-free expression images, and selects the K-means clustering centers with good data characteristics, which are used as the initial value of the convolution kernel of the CNN model to extract features. Secondly, using the feature extraction processing of the convolutional neural network, the extracted features are fed to the multi-class SVM classifier. The experimental results show that the proposed method reduces the training time of the model overall, improves the accuracies of facial expression recognition under the background of complex images, and has a certain robustness.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"92 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133888993","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}
Underwater sonar imaging system has been widely utilized to detect and identify the submerged objects of interest. However, imaging quality often suffers from the undesirable signal-dependent speckle noise during signal acquisition and transmission. The speckle noise will restrict the practical applications, such as object detection, tracking and recognition, etc. To enhance the sonar imaging performance, we propose a deep learning approach to directly estimate the speckle noise in logarithmic domain based on the convolutional neural network. Once the speckle noise is obtained, the latent sharp image can then be easily calculated according to the image degradation model. The patch-based loss function, i.e., structural similarity metric, is adopted to preserve the important geometrical structures during speckle noise reduction. Experiments have been implemented on different noise levels to demonstrate the effectiveness of the proposed deep learning approach. Experimental results have illustrated that it outperforms several widely-used speckle noise reduction methods.
{"title":"Learning a Deep Convolutional Network for Speckle Noise Reduction in Underwater Sonar Images","authors":"Yuxu Lu, R. W. Liu, Fenge Chen, Liang Xie","doi":"10.1145/3318299.3318358","DOIUrl":"https://doi.org/10.1145/3318299.3318358","url":null,"abstract":"Underwater sonar imaging system has been widely utilized to detect and identify the submerged objects of interest. However, imaging quality often suffers from the undesirable signal-dependent speckle noise during signal acquisition and transmission. The speckle noise will restrict the practical applications, such as object detection, tracking and recognition, etc. To enhance the sonar imaging performance, we propose a deep learning approach to directly estimate the speckle noise in logarithmic domain based on the convolutional neural network. Once the speckle noise is obtained, the latent sharp image can then be easily calculated according to the image degradation model. The patch-based loss function, i.e., structural similarity metric, is adopted to preserve the important geometrical structures during speckle noise reduction. Experiments have been implemented on different noise levels to demonstrate the effectiveness of the proposed deep learning approach. Experimental results have illustrated that it outperforms several widely-used speckle noise reduction methods.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130515940","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}
Jinyin Chen, Yitao Yang, Keke Hu, Hai-bin Zheng, Zhen Wang
With the continuous development of web services, the web security becomes more and more important. Distributed Denial of Service (DDoS) attack as one of the most common form of attacks, has produced serious economic damages. DDoS attack detection as one of main defense methods is suffered increasing attention by researchers. Most of them use machine learning methods to make good detection performance. However, there are still gaps between real detection rate and expected one, conventional machine learning methods are limited compared with deep learning. In this paper, we propose DAD-MCNN, a multi-channel CNN(MC-CNN) based DDoS attack detection framework, which can fully utilize information from a huge amount of network packages and set up an earlier warning system. Our contributions can be summarized as follows: (1) we propose a new preprocessing method for the network dataset. (2) MC-CNN is applied to detect DDoS attack and the detection result is decided by data in respective channels. (3) We use incremental training method to optimize training procedures and time in MC-CNN. (4) The experiment result shows that MC-CNN has the highest accuracy compared with conventional machine learning methods. The result also proves that our approach has performed well not only in DDoS attack detection but also in other anomaly attack detection.
{"title":"DAD-MCNN: DDoS Attack Detection via Multi-channel CNN","authors":"Jinyin Chen, Yitao Yang, Keke Hu, Hai-bin Zheng, Zhen Wang","doi":"10.1145/3318299.3318329","DOIUrl":"https://doi.org/10.1145/3318299.3318329","url":null,"abstract":"With the continuous development of web services, the web security becomes more and more important. Distributed Denial of Service (DDoS) attack as one of the most common form of attacks, has produced serious economic damages. DDoS attack detection as one of main defense methods is suffered increasing attention by researchers. Most of them use machine learning methods to make good detection performance. However, there are still gaps between real detection rate and expected one, conventional machine learning methods are limited compared with deep learning. In this paper, we propose DAD-MCNN, a multi-channel CNN(MC-CNN) based DDoS attack detection framework, which can fully utilize information from a huge amount of network packages and set up an earlier warning system. Our contributions can be summarized as follows: (1) we propose a new preprocessing method for the network dataset. (2) MC-CNN is applied to detect DDoS attack and the detection result is decided by data in respective channels. (3) We use incremental training method to optimize training procedures and time in MC-CNN. (4) The experiment result shows that MC-CNN has the highest accuracy compared with conventional machine learning methods. The result also proves that our approach has performed well not only in DDoS attack detection but also in other anomaly attack detection.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115209731","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}
Xingyu Fu, Bin Fang, Jiye Qian, Zhenni Wu, Jiajie Zhu, Tongxin Du
This paper presents an improved traffic sign detection method based on Faster R-CNN with dataset augmentation and subcategory detection scheme. Firstly, we extract natural scene frames from given videos and determine 20 categories of traffic signs. Secondly, we extend the image dataset and extract regions of interest, then manually annotate all categories. Thirdly, we train the Faster R-CNN model based on TensorFlow, then test the model and obtain the following evaluation indexes: the mean average precision is 99.07%, the recall rate is 99.66%, and the precision rate is 97.54%. Finally, we add the subcategory detection scheme to determine traffic light states, and we get the following evaluation indexes: the mean average precision is 99.50%, the recall rate is 100%, and the precision rate is 94.40%. Our experiments prove the robustness and accuracy for both traffic sign detection and subcategory detection of traffic light.
{"title":"Roadside Traffic Sign Detection Based on Faster R-CNN","authors":"Xingyu Fu, Bin Fang, Jiye Qian, Zhenni Wu, Jiajie Zhu, Tongxin Du","doi":"10.1145/3318299.3318348","DOIUrl":"https://doi.org/10.1145/3318299.3318348","url":null,"abstract":"This paper presents an improved traffic sign detection method based on Faster R-CNN with dataset augmentation and subcategory detection scheme. Firstly, we extract natural scene frames from given videos and determine 20 categories of traffic signs. Secondly, we extend the image dataset and extract regions of interest, then manually annotate all categories. Thirdly, we train the Faster R-CNN model based on TensorFlow, then test the model and obtain the following evaluation indexes: the mean average precision is 99.07%, the recall rate is 99.66%, and the precision rate is 97.54%. Finally, we add the subcategory detection scheme to determine traffic light states, and we get the following evaluation indexes: the mean average precision is 99.50%, the recall rate is 100%, and the precision rate is 94.40%. Our experiments prove the robustness and accuracy for both traffic sign detection and subcategory detection of traffic light.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121291336","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 this paper, we present a distributed machine learning algorithm over a network with fixed-delay tolerance. The network is directed and strongly connected. The training dataset is distributed to all agents in the network. We combine the distributed convex optimization (which utilizes double linear iterations) and corresponding machine learning algorithm. Each agent can only access its own local dataset. Suppose the delay between any pair of agents is time-invariant. The simulation shows that our algorithm is able to work under delayed transmission, in the sense that over time at each agent t the ratio of the estimate value xi(t) and scaling variable yi(t) can converge to the optimal point of the global cost function corresponding to the machine learning problem.
{"title":"Distributed Machine Learning over Directed Network with Fixed Communication Delays","authors":"Guo Zhenning","doi":"10.1145/3318299.3318340","DOIUrl":"https://doi.org/10.1145/3318299.3318340","url":null,"abstract":"In this paper, we present a distributed machine learning algorithm over a network with fixed-delay tolerance. The network is directed and strongly connected. The training dataset is distributed to all agents in the network. We combine the distributed convex optimization (which utilizes double linear iterations) and corresponding machine learning algorithm. Each agent can only access its own local dataset. Suppose the delay between any pair of agents is time-invariant. The simulation shows that our algorithm is able to work under delayed transmission, in the sense that over time at each agent t the ratio of the estimate value xi(t) and scaling variable yi(t) can converge to the optimal point of the global cost function corresponding to the machine learning problem.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121362752","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}
Personality analysis on social media is a research hotspot due to the importance of personality research in psychology as well as the rapid development of social media. Many studies have used social media status to analyze user's personality, but most of them are conducted on inadequate label data and linguistic features. In this paper, to explore the usage of unlabeled data on personality analysis, a personality analysis framework based on semi-supervised learning is introduced. Besides, for making full use of the language information in social media status, the well-known n-gram model is adopted to extract linguistic features. The experimental results demonstrate the semi-supervised learning can take advantage of unlabeled data and improve the accuracy of prediction model.
{"title":"Predicting Personality Using Facebook Status Based on Semi-supervised Learning","authors":"Heci Zheng, Chunhua Wu","doi":"10.1145/3318299.3318363","DOIUrl":"https://doi.org/10.1145/3318299.3318363","url":null,"abstract":"Personality analysis on social media is a research hotspot due to the importance of personality research in psychology as well as the rapid development of social media. Many studies have used social media status to analyze user's personality, but most of them are conducted on inadequate label data and linguistic features. In this paper, to explore the usage of unlabeled data on personality analysis, a personality analysis framework based on semi-supervised learning is introduced. Besides, for making full use of the language information in social media status, the well-known n-gram model is adopted to extract linguistic features. The experimental results demonstrate the semi-supervised learning can take advantage of unlabeled data and improve the accuracy of prediction model.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123568795","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}
As one of the most classical clustering techniques, the k-means clustering has been widely used in various areas over the past few decades. Despite its significant success, there are still several challenging issues in the k-means clustering research, one of which lies in its high sensitivity to the selection of the initial cluster centers. In this paper, we propose a new cluster center initialization method for k-means based on ensemble learning. Specifically, an ensemble of base clusterings are first constructed by using multiple k-means clusterers with random initializations. Then, a co-association matrix is computed for the base clusterings, upon which the agglomerative clustering algorithm can thereby be performed to build a pre-clustering result. From the pre-clustering, the set of initial cluster centers are obtained and then used for the final k-means clustering process. Experiments on multiple real-world datasets have demonstrated the superiority of the proposed method.
{"title":"Ensemble-Initialized k-Means Clustering","authors":"Shasha Xu, Dong Huang","doi":"10.1145/3318299.3318308","DOIUrl":"https://doi.org/10.1145/3318299.3318308","url":null,"abstract":"As one of the most classical clustering techniques, the k-means clustering has been widely used in various areas over the past few decades. Despite its significant success, there are still several challenging issues in the k-means clustering research, one of which lies in its high sensitivity to the selection of the initial cluster centers. In this paper, we propose a new cluster center initialization method for k-means based on ensemble learning. Specifically, an ensemble of base clusterings are first constructed by using multiple k-means clusterers with random initializations. Then, a co-association matrix is computed for the base clusterings, upon which the agglomerative clustering algorithm can thereby be performed to build a pre-clustering result. From the pre-clustering, the set of initial cluster centers are obtained and then used for the final k-means clustering process. Experiments on multiple real-world datasets have demonstrated the superiority of the proposed method.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124630921","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}
Machine learning has achieved outstanding performance in many fields, but its success heavily relies on the large number of annotated training samples. However, for many professional fields, data annotation is not only tedious and time consuming, but also demanding specialty-oriented knowledge and skills, which are not easily accessible. To significantly reduce the cost of annotation, we propose a novel active learning framework called ALBS. ALBS uses the syncretic strategy which incorporates "most discriminative" and "most representative" to seek "worthy" samples from unlabeled dataset and update the model incrementally to enhance the performance continuously. We have evaluated our method on two different audio datasets, demonstrating that the syncretic strategy can makes the promotion of model model's performance more robust and faster than the other strategies, and subsampling the historical labeled dataset can reduce unnecessary computing costs and storage space.
{"title":"ALBS: An Active Learning Framework Based on Syncretic Sample Selection Strategy","authors":"Longfei Pan, Xiaojun Wang","doi":"10.1145/3318299.3318362","DOIUrl":"https://doi.org/10.1145/3318299.3318362","url":null,"abstract":"Machine learning has achieved outstanding performance in many fields, but its success heavily relies on the large number of annotated training samples. However, for many professional fields, data annotation is not only tedious and time consuming, but also demanding specialty-oriented knowledge and skills, which are not easily accessible. To significantly reduce the cost of annotation, we propose a novel active learning framework called ALBS. ALBS uses the syncretic strategy which incorporates \"most discriminative\" and \"most representative\" to seek \"worthy\" samples from unlabeled dataset and update the model incrementally to enhance the performance continuously. We have evaluated our method on two different audio datasets, demonstrating that the syncretic strategy can makes the promotion of model model's performance more robust and faster than the other strategies, and subsampling the historical labeled dataset can reduce unnecessary computing costs and storage space.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116736660","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}
Chao Xia, Yawen Xiao, Jun Wu, Xiaodong Zhao, Hua Li
Cancer is a deadly disease all over the world and its morbidity is increasing at an alarming rate in recent years. With the rapid development of computer science and machine learning technologies, computer-aid cancer prediction has achieved increasingly progress. DNA methylation, as an important epigenetic modification, plays a vital role in the formation and progression of cancer, and therefore can be used as a feature for cancer identification. In this study, we introduce a convolutional neural network based multi-model ensemble method for cancer prediction using DNA methylation data. We first choose five basic machine learning methods as the first stage classifiers and conduct prediction individually. Then, a convolutional neural network is used to find the high-level features among the classifiers and gives a credible prediction result. Experimental results on three DNA methylation datasets of Lung Adenocarcinoma, Liver Hepatocellular Carcinoma and Kidney Clear Cell Carcinoma show the proposed ensemble method can uncover the intricate relationship among the classifiers automatically and achieve better performances.
{"title":"A Convolutional Neural Network Based Ensemble Method for Cancer Prediction Using DNA Methylation Data","authors":"Chao Xia, Yawen Xiao, Jun Wu, Xiaodong Zhao, Hua Li","doi":"10.1145/3318299.3318372","DOIUrl":"https://doi.org/10.1145/3318299.3318372","url":null,"abstract":"Cancer is a deadly disease all over the world and its morbidity is increasing at an alarming rate in recent years. With the rapid development of computer science and machine learning technologies, computer-aid cancer prediction has achieved increasingly progress. DNA methylation, as an important epigenetic modification, plays a vital role in the formation and progression of cancer, and therefore can be used as a feature for cancer identification. In this study, we introduce a convolutional neural network based multi-model ensemble method for cancer prediction using DNA methylation data. We first choose five basic machine learning methods as the first stage classifiers and conduct prediction individually. Then, a convolutional neural network is used to find the high-level features among the classifiers and gives a credible prediction result. Experimental results on three DNA methylation datasets of Lung Adenocarcinoma, Liver Hepatocellular Carcinoma and Kidney Clear Cell Carcinoma show the proposed ensemble method can uncover the intricate relationship among the classifiers automatically and achieve better performances.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117012563","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}
Iqra Tahir, Usman Qamar, Hassan Abbas, Babar Zeb, Sana Abid
EEG signal classification is a pivotal task for identification of different brain related disorders. The paper is about classification of EEG signal presenting a novel approach for the identification of whether the seizure is epileptic or normal that technique is based on training of neural network with having improved simplified swarm optimization algorithm. Our proposed methodology is evaluated with different parameters and testing accuracy of 94 % is reported for a publicly available dataset.
{"title":"Classification of EEG Signal by Training Neural Network with Swarm Optimization for Identification of Epilepsy","authors":"Iqra Tahir, Usman Qamar, Hassan Abbas, Babar Zeb, Sana Abid","doi":"10.1145/3318299.3318374","DOIUrl":"https://doi.org/10.1145/3318299.3318374","url":null,"abstract":"EEG signal classification is a pivotal task for identification of different brain related disorders. The paper is about classification of EEG signal presenting a novel approach for the identification of whether the seizure is epileptic or normal that technique is based on training of neural network with having improved simplified swarm optimization algorithm. Our proposed methodology is evaluated with different parameters and testing accuracy of 94 % is reported for a publicly available dataset.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"360 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121717901","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：