Hashing algorithm projects the vector of features onto the binary space that generate the binary codes to reduce calculating time. Thus Hashing Algorithm is widely used to improve retrieval efficiency in traditional image retrieval methods based on Deep neural networks (DNNs). In this paper, we extract the feature vectors whose elements between 0 and 1 by DNNs and linear scaling method, then we define the mean of each column vector of the matrix consisted of these feature vectors as threshold to create corresponding hashing codes after two-stages binarization. Since threshold brings major effect to the preservation of the similarity between images, during this process, the two-stages binarization play two important roles: 1) optimizing thresholds; 2) optimizing hash codes. The promising experimental results on public available Cifar-10 database show that the proposed approach achieve higher precision compared with the state-of-the-art hashing algorithms.
{"title":"An Improved Hashing Method for Image Retrieval Based on Deep Neural Networks","authors":"Qiu Chen, Weidong Wang, Feifei Lee","doi":"10.1145/3297067.3297092","DOIUrl":"https://doi.org/10.1145/3297067.3297092","url":null,"abstract":"Hashing algorithm projects the vector of features onto the binary space that generate the binary codes to reduce calculating time. Thus Hashing Algorithm is widely used to improve retrieval efficiency in traditional image retrieval methods based on Deep neural networks (DNNs). In this paper, we extract the feature vectors whose elements between 0 and 1 by DNNs and linear scaling method, then we define the mean of each column vector of the matrix consisted of these feature vectors as threshold to create corresponding hashing codes after two-stages binarization. Since threshold brings major effect to the preservation of the similarity between images, during this process, the two-stages binarization play two important roles: 1) optimizing thresholds; 2) optimizing hash codes. The promising experimental results on public available Cifar-10 database show that the proposed approach achieve higher precision compared with the state-of-the-art hashing algorithms.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129042444","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}
Based on instrumental sensors, a cascade calibration method of the near-field source is proposed. The method can not only uses multiple independent near-field signals operating at different times and different locations calibrate the gain and phase errors and position errors, but also locate the near-field source at the same time. At the single signal, reconstructing the virtual array and steering vector transformation are taken. Compared to the joint estimation of multidimensional parameters, the method can be estimated in real time and less affected by error variations. Only one-dimensional spectral search is needed and there is no loss of aperture in constructing the virtual array. Simultaneously, simulation experiments show the performance of the proposed algorithm in this paper.
{"title":"A Cascade Method for Two Kinds of Errors Calibration in Array","authors":"Meng-yu Ni, Song Xiao, Hui Chen, Longxiang Li","doi":"10.1145/3297067.3297084","DOIUrl":"https://doi.org/10.1145/3297067.3297084","url":null,"abstract":"Based on instrumental sensors, a cascade calibration method of the near-field source is proposed. The method can not only uses multiple independent near-field signals operating at different times and different locations calibrate the gain and phase errors and position errors, but also locate the near-field source at the same time. At the single signal, reconstructing the virtual array and steering vector transformation are taken. Compared to the joint estimation of multidimensional parameters, the method can be estimated in real time and less affected by error variations. Only one-dimensional spectral search is needed and there is no loss of aperture in constructing the virtual array. Simultaneously, simulation experiments show the performance of the proposed algorithm in this paper.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"90 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122006451","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}
Arabic topic detection (ATD) has become an attractive research field. It is used in many applications, such as Arabic documents classification, web search, social media, and security. ATD uses machine learning algorithms with ultimate aim to classify Arabic documents based on text contents. Arabic text classification require a complicated process. The Arabic words have unlimited variation in the meaning, which add more complexity and ambiguity to the process Arabic text classification. There are some studies have been proposed for Arabic text classification in recent years. However, these previous studies need improvements to rise accuracy and efficiency. Therefore, this paper proposes an effective approach for Arabic text classification and topic detection using discriminative multi nominal naïve Bayes (DMNB) classifier and frequency transform. The proposed approach includes three main steps: Arabic text preprocessing, Arabic text feature extraction and normalization, and Arabic text classification. A dataset of 1500 Arabic documents collected from Arabic articles corpus in 5 different topics is used to evaluate the proposed approach. The experimental results of 10-folds cross-validation show that the proposed approach performs competitively better than the state-of-the-art approaches.
{"title":"Arabic Topic Detection Using Discriminative Multi nominal Naïve Bayes and Frequency Transforms","authors":"Ahmed Alsanad","doi":"10.1145/3297067.3297095","DOIUrl":"https://doi.org/10.1145/3297067.3297095","url":null,"abstract":"Arabic topic detection (ATD) has become an attractive research field. It is used in many applications, such as Arabic documents classification, web search, social media, and security. ATD uses machine learning algorithms with ultimate aim to classify Arabic documents based on text contents. Arabic text classification require a complicated process. The Arabic words have unlimited variation in the meaning, which add more complexity and ambiguity to the process Arabic text classification. There are some studies have been proposed for Arabic text classification in recent years. However, these previous studies need improvements to rise accuracy and efficiency. Therefore, this paper proposes an effective approach for Arabic text classification and topic detection using discriminative multi nominal naïve Bayes (DMNB) classifier and frequency transform. The proposed approach includes three main steps: Arabic text preprocessing, Arabic text feature extraction and normalization, and Arabic text classification. A dataset of 1500 Arabic documents collected from Arabic articles corpus in 5 different topics is used to evaluate the proposed approach. The experimental results of 10-folds cross-validation show that the proposed approach performs competitively better than the state-of-the-art approaches.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"184 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124086270","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 ability of software systems adapting to human's input is a key element in the symbiosis of human-system co-adaptation, where human and software-based systems work together in a close partnership to achieve synergetic goals. This seamless integration will eliminate the barriers between human and machine. A critical requirement for co-adaptive systems is software system's ability to recognize human emotion, in which the system can detect and interpret users' emotions and adapt accordingly. There are numerous solutions that provide the technologies for emotion recognition. However, selecting an appropriate solution for a given task within a specific application domain can be challenging. The vast variation between these solutions makes the selecting task even more difficult. This paper compares cloud-based emotion recognition services offered by Amazon, Google, and Microsoft. These services detect human emotion through facial expression recognition with the utilization of computer vision. The focus of this paper is to measure the detection accuracy of these services. Accuracy is calculated based on the highest confidence rating returned by each service. All three services have been tested with the same dataset. This paper concludes with findings and recommendations based on our comparative analysis among these services.
{"title":"A Comparative Study on Detection Accuracy of Cloud-Based Emotion Recognition Services","authors":"Osamah M. Al-Omair, Shihong Huang","doi":"10.1145/3297067.3297079","DOIUrl":"https://doi.org/10.1145/3297067.3297079","url":null,"abstract":"The ability of software systems adapting to human's input is a key element in the symbiosis of human-system co-adaptation, where human and software-based systems work together in a close partnership to achieve synergetic goals. This seamless integration will eliminate the barriers between human and machine. A critical requirement for co-adaptive systems is software system's ability to recognize human emotion, in which the system can detect and interpret users' emotions and adapt accordingly. There are numerous solutions that provide the technologies for emotion recognition. However, selecting an appropriate solution for a given task within a specific application domain can be challenging. The vast variation between these solutions makes the selecting task even more difficult. This paper compares cloud-based emotion recognition services offered by Amazon, Google, and Microsoft. These services detect human emotion through facial expression recognition with the utilization of computer vision. The focus of this paper is to measure the detection accuracy of these services. Accuracy is calculated based on the highest confidence rating returned by each service. All three services have been tested with the same dataset. This paper concludes with findings and recommendations based on our comparative analysis among these services.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"283 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121824443","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}
Online transfer learning aims to attack an online learning task on a target domain by transferring knowledge from some source domains, which has received more attentions. And most online transfer learning methods adapt the classifier according to its accuracy on new coming data. However, in real-world applications, such as anomaly detection and credit card fraud detection, the cost may be more important than the accuracy. Moreover, the cost usually changes in these online data, which challenges state-of-art-methods. Therefore, this paper introduces the cost of misclassification into transfer-learning of classifier, and proposes a novel online transfer learning algorithm with adaptive cost (OLAC). Firstly, we introduce the label distribution into traditional Hinge Loss Function to compute the cost of classification adaptively. Secondly, we transfer learn the classifier according to its performance on new coming data including both accuracy and cost. Extensive experimental results show that our method can achieve higher accuracy and less classification lost, especially for the samples with higher costs.
在线迁移学习的目的是通过对源领域的知识进行迁移，在目标领域上完成在线学习任务，近年来受到越来越多的关注。大多数在线迁移学习方法都是根据分类器对新数据的准确性来调整分类器的。然而，在实际应用中，例如异常检测和信用卡欺诈检测，成本可能比准确性更重要。此外，这些在线数据的成本通常会发生变化，这对最先进的方法提出了挑战。为此，本文将误分类代价引入分类器迁移学习中，提出了一种具有自适应代价(OLAC)的在线迁移学习算法。首先，在传统的Hinge Loss Function中引入标签分布，自适应计算分类代价;其次，我们根据分类器对新数据的处理性能，包括准确率和成本，对分类器进行迁移学习。大量的实验结果表明，我们的方法可以达到更高的准确率和更少的分类损失，特别是对于成本较高的样本。
{"title":"An Online Transfer Learning Algorithm with Adaptive Cost","authors":"Yuhong Zhang, Mimi Wu, Xuegang Hu, Yi Zhu","doi":"10.1145/3297067.3297075","DOIUrl":"https://doi.org/10.1145/3297067.3297075","url":null,"abstract":"Online transfer learning aims to attack an online learning task on a target domain by transferring knowledge from some source domains, which has received more attentions. And most online transfer learning methods adapt the classifier according to its accuracy on new coming data. However, in real-world applications, such as anomaly detection and credit card fraud detection, the cost may be more important than the accuracy. Moreover, the cost usually changes in these online data, which challenges state-of-art-methods. Therefore, this paper introduces the cost of misclassification into transfer-learning of classifier, and proposes a novel online transfer learning algorithm with adaptive cost (OLAC). Firstly, we introduce the label distribution into traditional Hinge Loss Function to compute the cost of classification adaptively. Secondly, we transfer learn the classifier according to its performance on new coming data including both accuracy and cost. Extensive experimental results show that our method can achieve higher accuracy and less classification lost, especially for the samples with higher costs.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122049820","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}
Hiroki Yamamoto, Kazunori Iwata, N. Suematsu, Kazushi Mimura
Regarding shape matching, we present a novel method of determining a correspondence between shapes that is applicable to existing local descriptors and somewhat enhances them. In our method, we determine the correspondence of a focusing point of a shape, considering the correspondence of neighboring points to the focusing point. This plays a vital role in avoiding the risk of failing to notice a more appropriate correspondence. However, considering neighboring points causes another problem of computational feasibility because there is a considerable increase in the number of possible correspondences searched in matching shapes. We therefore manage this problem using an efficient approximation to reduce the number of possible correspondences. Conducting numerical analysis on shape retrieval, we show that our method is useful for obtaining a better correspondence than the conventional method that does not consider the correspondence of neighboring points.
{"title":"A Shape Matching Method Considering Computational Feasibility","authors":"Hiroki Yamamoto, Kazunori Iwata, N. Suematsu, Kazushi Mimura","doi":"10.1145/3297067.3297077","DOIUrl":"https://doi.org/10.1145/3297067.3297077","url":null,"abstract":"Regarding shape matching, we present a novel method of determining a correspondence between shapes that is applicable to existing local descriptors and somewhat enhances them. In our method, we determine the correspondence of a focusing point of a shape, considering the correspondence of neighboring points to the focusing point. This plays a vital role in avoiding the risk of failing to notice a more appropriate correspondence. However, considering neighboring points causes another problem of computational feasibility because there is a considerable increase in the number of possible correspondences searched in matching shapes. We therefore manage this problem using an efficient approximation to reduce the number of possible correspondences. Conducting numerical analysis on shape retrieval, we show that our method is useful for obtaining a better correspondence than the conventional method that does not consider the correspondence of neighboring points.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128581830","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}
With the advantage of observing the earth all the day, the Synthetic Aperture Radar (SAR) has become an important means to image the interest targets. However, the traditional SAR is still large and has the characteristics of large volume and high resource cost. In recent years, the vigorous development of micro satellite field has also promoted the research of small, intelligent and distributed cooperative imaging field of space borne SAR. The traditional spaceborne SAR imaging method is to process the echo of the received LFM signal in the two-dimensional pulse compression domain. Since the targets in the ocean are usually sparsed, it is an effective method to detect the interested targets by analyzing the echo and transform domain. In this paper, the method of azimuth pulse compression is used to realize the perception and fast imaging of sparse targets. Compared with the traditional method, it's clear that the method has a greater improvement in resource and time consumption performance.
{"title":"Fast Target Perception Imaging of Spaceborne SAR in Sparse Field","authors":"Pan Zhang, Yi Huang, Zhonghe Jin","doi":"10.1145/3432291.3432301","DOIUrl":"https://doi.org/10.1145/3432291.3432301","url":null,"abstract":"With the advantage of observing the earth all the day, the Synthetic Aperture Radar (SAR) has become an important means to image the interest targets. However, the traditional SAR is still large and has the characteristics of large volume and high resource cost. In recent years, the vigorous development of micro satellite field has also promoted the research of small, intelligent and distributed cooperative imaging field of space borne SAR. The traditional spaceborne SAR imaging method is to process the echo of the received LFM signal in the two-dimensional pulse compression domain. Since the targets in the ocean are usually sparsed, it is an effective method to detect the interested targets by analyzing the echo and transform domain. In this paper, the method of azimuth pulse compression is used to realize the perception and fast imaging of sparse targets. Compared with the traditional method, it's clear that the method has a greater improvement in resource and time consumption performance.","PeriodicalId":340004,"journal":{"name":"International Conference on Signal Processing and Machine Learning","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122062841","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：