Dedong Tang, Chen Li, Xiaohui Ji, Zhenyu Chen, Fangchun Di
The power load forecasting is based on historical energy consumption data of a region to forecast the power consumption of the region for a period of time in the future. Accurate forecasting can provide effective and reliable guidance for power construction and grid operation. This paper proposed a power load forecasting approach using a two LSTM (long-short-term memory) layers neural network. Based on the real power load data provided by EUNITE, a power load forecasting method based on LSTM is constructed. Two models, single-point forecasting model and multiple-point forecasting model, are built to forecast the power of next hour and next half day. The experimental results show that the mean absolute percentage error of the single-point forecasting model is 1.806 and the mean absolute percentage error of the multiple-points forecasting model of LSTM network is 2.496.
{"title":"Power Load Forecasting Using a Refined LSTM","authors":"Dedong Tang, Chen Li, Xiaohui Ji, Zhenyu Chen, Fangchun Di","doi":"10.1145/3318299.3318353","DOIUrl":"https://doi.org/10.1145/3318299.3318353","url":null,"abstract":"The power load forecasting is based on historical energy consumption data of a region to forecast the power consumption of the region for a period of time in the future. Accurate forecasting can provide effective and reliable guidance for power construction and grid operation. This paper proposed a power load forecasting approach using a two LSTM (long-short-term memory) layers neural network. Based on the real power load data provided by EUNITE, a power load forecasting method based on LSTM is constructed. Two models, single-point forecasting model and multiple-point forecasting model, are built to forecast the power of next hour and next half day. The experimental results show that the mean absolute percentage error of the single-point forecasting model is 1.806 and the mean absolute percentage error of the multiple-points forecasting model of LSTM network is 2.496.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"28 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":"114671837","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}
Convolutional neural networks (CNNs) are extremely important building blocks for abstract deep learning algorithm constructs regarding visual interpretation especially when it comes to synthetic aperture radar (SAR) images. An ongoing research is being made in order to improve their accuracy forgetting about the undiscovered internals. CNNs are usually being used as black boxes that produce in a non-linear fashion abstract interpretations. In this paper, however, we propose a novel algorithm that shows where CNNs look in an image to provide the answer to the provided classification problem applied to SAR images. We provide also results as bounding boxes using only a pre-trained classification network and some post-processing. The algorithm uses a brute-force approach given a pre-trained neural network, it removes gradually lines of pixels and checks the effect on the resulting scores, and it post-processes the resulting scores to infer the most important region in a given input image. Although other attempts have been made in the literature to provide solutions to the problem, by reversing the convolutional map filters, they are limited in scope and generally fail to deal with a complex network such as the award winning Resnet. Our algorithm, in this category, is of significant usefulness, it bridges the gap between the object classification and object detection problems, opening new perspectives to eliminate the time-consuming task of manual object annotation.
{"title":"A Novel Single Target Auto-annotation Algorithm for SAR Images Based on Pre-trained CNN Classifier","authors":"Moulay Idriss Bellil, Xiaojian Xu","doi":"10.1145/3318299.3318366","DOIUrl":"https://doi.org/10.1145/3318299.3318366","url":null,"abstract":"Convolutional neural networks (CNNs) are extremely important building blocks for abstract deep learning algorithm constructs regarding visual interpretation especially when it comes to synthetic aperture radar (SAR) images. An ongoing research is being made in order to improve their accuracy forgetting about the undiscovered internals. CNNs are usually being used as black boxes that produce in a non-linear fashion abstract interpretations. In this paper, however, we propose a novel algorithm that shows where CNNs look in an image to provide the answer to the provided classification problem applied to SAR images. We provide also results as bounding boxes using only a pre-trained classification network and some post-processing. The algorithm uses a brute-force approach given a pre-trained neural network, it removes gradually lines of pixels and checks the effect on the resulting scores, and it post-processes the resulting scores to infer the most important region in a given input image. Although other attempts have been made in the literature to provide solutions to the problem, by reversing the convolutional map filters, they are limited in scope and generally fail to deal with a complex network such as the award winning Resnet. Our algorithm, in this category, is of significant usefulness, it bridges the gap between the object classification and object detection problems, opening new perspectives to eliminate the time-consuming task of manual object annotation.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"726 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":"122999564","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}
It is difficult to discern the authenticity online reviews, which is critical particularly in a setting of patient-doctor online forum. In this paper, a model on the detection of doctor quality has been developed and tested with online big data. In this study, a database with 31,646 online reviews was compiled. Text mining and word-cloud analysis results indicate that the model provides an effective solution to assess the quality of doctors registered in online forum, the quality of doctor-patient online interaction, and patients' overall perception. A guideline has been provided to evaluate doctor authenticity.
{"title":"Online Forum Authenticity: Big Data Analytics in Healthcare","authors":"G. Zhan","doi":"10.1145/3318299.3318395","DOIUrl":"https://doi.org/10.1145/3318299.3318395","url":null,"abstract":"It is difficult to discern the authenticity online reviews, which is critical particularly in a setting of patient-doctor online forum. In this paper, a model on the detection of doctor quality has been developed and tested with online big data. In this study, a database with 31,646 online reviews was compiled. Text mining and word-cloud analysis results indicate that the model provides an effective solution to assess the quality of doctors registered in online forum, the quality of doctor-patient online interaction, and patients' overall perception. A guideline has been provided to evaluate doctor authenticity.","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":"124593817","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 proposed a weighted KNN algorithm based on random forests. The proposed algorithm fully measures the differences in the importance of each feature, and overcomes the shortcoming of k-nearest neighbor (KNN) algorithm in classifying unbalanced data sets and data sets of different feature importance. The classification accuracy of the KNN algorithm is effectively improved, and the performance of the proposed algorithm is verified through experiments.
{"title":"Weighted KNN Algorithm Based on Random Forests","authors":"Huanian Zhang, Fanliang Bu","doi":"10.1145/3318299.3318313","DOIUrl":"https://doi.org/10.1145/3318299.3318313","url":null,"abstract":"In this paper, we proposed a weighted KNN algorithm based on random forests. The proposed algorithm fully measures the differences in the importance of each feature, and overcomes the shortcoming of k-nearest neighbor (KNN) algorithm in classifying unbalanced data sets and data sets of different feature importance. The classification accuracy of the KNN algorithm is effectively improved, and the performance of the proposed algorithm is verified through experiments.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"60 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":"130301728","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}
Mengmeng Chen, Lifen Jiang, Chunmei Ma, Huazhi Sun
Computer emotion recognition plays an important role in the field of artificial intelligence and is a key technology to realize human-machine interaction. Aiming at a cross-modal fusion problem of two nonlinear features of facial expression image and speech emotion, a bimodal fusion emotion recognition model (D-CNN) based on convolutional neural network is proposed. Firstly, a fine-grained feature extraction method based on convolutional neural network is proposed. Secondly, in order to obtain joint features representation, a feature fusion method based on the fine-grained features of bimodal is proposed. Finally, in order to verify the performance of the D-CNN model, experiments were conducted on the open source dataset eNTERFACE'05. The experimental results show that the multi-modal emotion recognition model D-CNN is more than 10% higher than the single emotion recognition model of speech and facial expression respectively. In addition, compared with the other commonly used bimodal emotion recognition methods(such as universal background model), the recognition rete of D-CNN is increased by 5%.
{"title":"Bimodal Emotion Recognition Based on Convolutional Neural Network","authors":"Mengmeng Chen, Lifen Jiang, Chunmei Ma, Huazhi Sun","doi":"10.1145/3318299.3318347","DOIUrl":"https://doi.org/10.1145/3318299.3318347","url":null,"abstract":"Computer emotion recognition plays an important role in the field of artificial intelligence and is a key technology to realize human-machine interaction. Aiming at a cross-modal fusion problem of two nonlinear features of facial expression image and speech emotion, a bimodal fusion emotion recognition model (D-CNN) based on convolutional neural network is proposed. Firstly, a fine-grained feature extraction method based on convolutional neural network is proposed. Secondly, in order to obtain joint features representation, a feature fusion method based on the fine-grained features of bimodal is proposed. Finally, in order to verify the performance of the D-CNN model, experiments were conducted on the open source dataset eNTERFACE'05. The experimental results show that the multi-modal emotion recognition model D-CNN is more than 10% higher than the single emotion recognition model of speech and facial expression respectively. In addition, compared with the other commonly used bimodal emotion recognition methods(such as universal background model), the recognition rete of D-CNN is increased by 5%.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"6 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":"126935217","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}
Most scene text detection methods based on deep learning are difficult to locate texts with multi-scale shapes. The challenges of scale robust text detection lie in two aspects: 1) scene text can be diverse and usually exists in various colors, fonts, orientations, languages, and scales in natural images. 2) Most existing detectors are difficult to locate text with large scale change. We propose a new Inception-Text module and adaptive scale scaling test mechanism for multi-oriented scene text detection. the proposed algorithm enhances performance significantly, while adding little computation. The proposed method can flexibly detect text in various scales, including horizontal, oriented and curved text. The proposed algorithm is evaluated on three recent standard public benchmarks, and show that our proposed method achieves the state-of-the-art performance on several benchmarks. Specifically, it achieves an F-measure of 93.3% on ICDAR2013, 90.47% on ICDAR2015 and 76.08%1 on ICDAR2017 MLT.
{"title":"Scene Text Detection with Inception Text Proposal Generation Module","authors":"Hang Zhang, Jiahang Liu, Tieqiao Chen","doi":"10.1145/3318299.3318373","DOIUrl":"https://doi.org/10.1145/3318299.3318373","url":null,"abstract":"Most scene text detection methods based on deep learning are difficult to locate texts with multi-scale shapes. The challenges of scale robust text detection lie in two aspects: 1) scene text can be diverse and usually exists in various colors, fonts, orientations, languages, and scales in natural images. 2) Most existing detectors are difficult to locate text with large scale change. We propose a new Inception-Text module and adaptive scale scaling test mechanism for multi-oriented scene text detection. the proposed algorithm enhances performance significantly, while adding little computation. The proposed method can flexibly detect text in various scales, including horizontal, oriented and curved text. The proposed algorithm is evaluated on three recent standard public benchmarks, and show that our proposed method achieves the state-of-the-art performance on several benchmarks. Specifically, it achieves an F-measure of 93.3% on ICDAR2013, 90.47% on ICDAR2015 and 76.08%1 on ICDAR2017 MLT.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"36 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":"130706837","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}
Drug-drug interactions (DDIs) can trigger unexpected pharmacological effects, including adverse drug events (ADEs), with causal mechanisms often unknown. Recently, deep neural network (DNN) models have achieved great success in many applications, including predicting pharmacological properties of drugs and drug repurposing. In this study, we generated features produced by SMILES (simplified molecular-input line-entry system) codes for more than 5,000 drugs downloaded from DrugBank. We built a deep neural network model to predict 80 DDI types using the features. We reached an overall accuracy and AUC (area under the curve) of receiver operating characteristic of 93.2% and 94.2% of the test data set and 94.9% and 95.6% of the validation data set, respectively. The trained model was applied to predict the DDI types of 13,155,885 drug-drug pairs combined by 5,130 drugs. The prediction results were applied to analyze the drugs currently used for treating inflammatory bowel disease (IBD). The potential drug combinations for treating IBD were discussed. These results can provide important insights on drug repurposing and guidelines during drug development.
{"title":"Predicting Drug-Drug Interactions Using Deep Neural Network","authors":"Xinyu Hou, Jiaying You, P. Hu","doi":"10.1145/3318299.3318323","DOIUrl":"https://doi.org/10.1145/3318299.3318323","url":null,"abstract":"Drug-drug interactions (DDIs) can trigger unexpected pharmacological effects, including adverse drug events (ADEs), with causal mechanisms often unknown. Recently, deep neural network (DNN) models have achieved great success in many applications, including predicting pharmacological properties of drugs and drug repurposing. In this study, we generated features produced by SMILES (simplified molecular-input line-entry system) codes for more than 5,000 drugs downloaded from DrugBank. We built a deep neural network model to predict 80 DDI types using the features. We reached an overall accuracy and AUC (area under the curve) of receiver operating characteristic of 93.2% and 94.2% of the test data set and 94.9% and 95.6% of the validation data set, respectively. The trained model was applied to predict the DDI types of 13,155,885 drug-drug pairs combined by 5,130 drugs. The prediction results were applied to analyze the drugs currently used for treating inflammatory bowel disease (IBD). The potential drug combinations for treating IBD were discussed. These results can provide important insights on drug repurposing and guidelines during drug development.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"21 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":"133199574","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}
LiangQi Zhou, Hongzhen Xu, Li Wei, Quan Zhang, Fei Zhou, Zhuo-Dai Li
Air quality has always been a hot issue of concern to the people, the environmental protection department and the government. Among the massive air quality data, abnormal data can interfere with subsequent experiments and analysis. Therefore, it is necessary to detect abnormal data to improve the accuracy of the data. However, traditional air outlier detection methods require at least one year's data to make inferences about air quality. This paper firstly analyzes the characteristics of air quality big data, and then proposes a framework based on Bayesian non-parametric clustering, namely Dirichlet Process (DP) clustering framework, to realize the outlier detection of air quality. The framework optimizes Gaussian mixture model into infinite Gaussian mixture model according to the results of data analysis, and uses neural network to cluster the data processed by infinite Gaussian mixture model, which effectively improves the clustering accuracy and avoids the need of collecting a large number of training data.
{"title":"Air Big Data Outlier Detection Based on Infinite Gauss Bayesian and CNN","authors":"LiangQi Zhou, Hongzhen Xu, Li Wei, Quan Zhang, Fei Zhou, Zhuo-Dai Li","doi":"10.1145/3318299.3318384","DOIUrl":"https://doi.org/10.1145/3318299.3318384","url":null,"abstract":"Air quality has always been a hot issue of concern to the people, the environmental protection department and the government. Among the massive air quality data, abnormal data can interfere with subsequent experiments and analysis. Therefore, it is necessary to detect abnormal data to improve the accuracy of the data. However, traditional air outlier detection methods require at least one year's data to make inferences about air quality. This paper firstly analyzes the characteristics of air quality big data, and then proposes a framework based on Bayesian non-parametric clustering, namely Dirichlet Process (DP) clustering framework, to realize the outlier detection of air quality. The framework optimizes Gaussian mixture model into infinite Gaussian mixture model according to the results of data analysis, and uses neural network to cluster the data processed by infinite Gaussian mixture model, which effectively improves the clustering accuracy and avoids the need of collecting a large number of training data.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"2 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":"128459956","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}
We propose a method to solve the problem of face replacing for image and video. This approach is enabled to transform an input identity into a target identity, including the facial expression, facial organs and the facial skin colour. To this end, we make the following contributions. (a)We elaborately design a simple auto encoder network to reconstruct the face. (b)Building on recent research in this area, we integrate a weight mask into the loss function to improve the performance of the network during training. (c)Unlike the previous work, we can transform the face not only in image, but also merging video after we adjust the results. We make it easier to replace a people's face with another one in image or video by combining neural networks with simple processing steps.
{"title":"A Face Replacement Neural Network for Image and Video","authors":"Yanhui Guo, Xue Ke, Jie Ma","doi":"10.1145/3318299.3318311","DOIUrl":"https://doi.org/10.1145/3318299.3318311","url":null,"abstract":"We propose a method to solve the problem of face replacing for image and video. This approach is enabled to transform an input identity into a target identity, including the facial expression, facial organs and the facial skin colour. To this end, we make the following contributions. (a)We elaborately design a simple auto encoder network to reconstruct the face. (b)Building on recent research in this area, we integrate a weight mask into the loss function to improve the performance of the network during training. (c)Unlike the previous work, we can transform the face not only in image, but also merging video after we adjust the results. We make it easier to replace a people's face with another one in image or video by combining neural networks with simple processing steps.","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":"129332634","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 Vector of Locally Aggregated Descriptors (VLAD) method, developed from BOW and Fisher Vector, has got great successes in image classification and retrieval. However, the traditional VLAD only assigns local descriptors to the closest visual words in the codebook, which is a hard voting process that leads to a large quantization error. In this paper, we propose an approach to fuse VLAD and locality-constrained linear coding (LLC), compared with the original method, several nearest neighbor centers are considered when assigning local descriptors. We use the reconstruction coefficients of LLC to obtain the weights of several nearest neighbor centers. Due to the excellent representation ability of the reconstruction coefficients for local descriptors, we also combine it with VLAD coding. Experiments were conducted on the 15 Scenes, UIUC Sports Event and Corel 10 datasets to demonstrate that our proposed method has outstanding performance in terms of classification accuracy. Our approach also does not generate much additional computational cost while encoding features.
{"title":"VLAD Encoding Based on LLC for Image Classification","authors":"Cheng Cheng, Xianzhong Long, Yun Li","doi":"10.1145/3318299.3318322","DOIUrl":"https://doi.org/10.1145/3318299.3318322","url":null,"abstract":"The Vector of Locally Aggregated Descriptors (VLAD) method, developed from BOW and Fisher Vector, has got great successes in image classification and retrieval. However, the traditional VLAD only assigns local descriptors to the closest visual words in the codebook, which is a hard voting process that leads to a large quantization error. In this paper, we propose an approach to fuse VLAD and locality-constrained linear coding (LLC), compared with the original method, several nearest neighbor centers are considered when assigning local descriptors. We use the reconstruction coefficients of LLC to obtain the weights of several nearest neighbor centers. Due to the excellent representation ability of the reconstruction coefficients for local descriptors, we also combine it with VLAD coding. Experiments were conducted on the 15 Scenes, UIUC Sports Event and Corel 10 datasets to demonstrate that our proposed method has outstanding performance in terms of classification accuracy. Our approach also does not generate much additional computational cost while encoding features.","PeriodicalId":164987,"journal":{"name":"International Conference on Machine Learning and Computing","volume":"46 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":"127063250","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：