检索结果-内蒙古大学图书馆

Computational aesthetics of photos quality assessment based on improved artificial neural network combined with an autoencoder technique

引用

NEUROCOMPUTING 2016年 188卷 50-62页

作者： Tan, Yunlan Zhou, Yimin Li, Guangyao Huang, Anmin Jinggangshan Univ Sch Elect & Informat Engn Jian Jiangxi Peoples R China Tongji Univ Coll Elect & Informat Engn Shanghai 200092 Peoples R China

Photograph aesthetical evaluation has been widely investigated in these decades. For fine-granularity aesthetic quality prediction, a novel aesthetics classifier based on improved artificial neural network combined with an autoencoder technique is presented. First, we download large consumer photographic images from a well-known online photograph portal. Then, we extract 56 features normalized to 0-1 and train the networks with photographs of high and low ratings to test the quality of photos. Experimental results show that the accuracy of classification is above 86.67%, which is better than all state-of-the-art methods. Meanwhile, it is observed from experiments that the extracted features are consistent with the humans' visual perception systems. (C) 2015 Elsevier B.V. All rights reserved.

关键词： Computational aesthetics Quality assessment Artificial neural network autoencoder High aesthetics Low aesthetics

来源：评论

学校读者我要写书评

暂无评论

Hyperspectral Unmixing Using Orthogonal Sparse Prior-Based autoencoder With Hyper-Laplacian Loss and Data-Driven Outlier Detection

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2020年第9期58卷 6550-6564页

作者： Dou, Zeyang Gao, Kun Zhang, Xiaodian Wang, Hong Wang, Junwei Beijing Inst Technol Minist Educ China Key Lab Photoelect Imaging Technol & Syst Beijing 100081 Peoples R China

Hyperspectral unmixing, which estimates end-members and their corresponding abundance fractions simultaneously, is an important task for hyperspectral applications. In this article, we propose a new autoencoder-based hyperspectral unmixing model with three novel components. First, we propose a new sparse prior to abundance maps. The proposed prior, called orthogonal sparse prior (OSP), is based on the observations that different abundance maps are close to orthogonal because, generally, no more than two end-members are mixed within one pixel. As opposed to the conventional norm-based sparse prior that assumes the abundance maps are independent, the proposed OSP explores the orthogonality between the abundance maps. Second, we propose the hyper-Laplacian loss to model the reconstruction error. The key observation is that the reconstruction error distribution usually has a heavy-tailed shape, which is better modeled by the hyper-Laplacian distribution rather than the commonly used Gaussian distribution. Third, to ease the side effect of outliers for end-member initializations, we develop a data-driven approach to detect outliers from the raw hyperspectral images. Extensive experiments on both synthetic and real-world data sets show that the proposed method significantly and consistently outperforms the compared state-of-the-art methods, with up to more than 50% improvements.

关键词： Hyperspectral imaging Decoding Image reconstruction Gaussian distribution Spatial resolution Anomaly detection autoencoder hyper-Laplacian distribution outlier detection sparse prior spectral unmixing

来源：评论

学校读者我要写书评

暂无评论

OIAE: Overall Improved autoencoder with Powerful Image Reconstruction and Discriminative Feature Extraction

引用

COGNITIVE COMPUTATION 2023年第4期15卷 1334-1341页

作者： Zhao, Haifeng Wu, Haiting Wang, Xin Anhui Univ Sch Comp Sci & Technol Minist Educ Key Lab Intelligent Comp & Signal Proc Shenzhen Peoples R China PKU Shenzhen Inst Shenzhen Peoples R China Anhui Univ Sch Comp Sci & Technol Hefei Peoples R China Shenzhen Raixun Informat Technol Co Ltd PKU Shenzhen Inst Shenzhen Peoples R China

As an unsupervised learning method, the autoencoder (AE) plays a very important role in model pre-training. However, the current AEs pre-training methods are still faced with the problems of not being able to reconstruct pictures better and mining deeper features. In this paper, we come up with a new AE, overall improved autoencoder (OIAE). Its main contribution is twofold: Wasserstein Generative Adversarial Networks (WGAN) is used to study the relationship between AEs reconstruction ability and pre-training performance and a regularization method is proposed to enable the autoencoder to learn discriminative features. We set up ablation experiments to prove the effectiveness of our two improvements and OIAE and compare them with baseline. The classification accuracy of the OIAE pre-trained classification network improved by 0.74% on the basic dataset and 16.44% on the more difficult dataset. These promising results demonstrate the effectiveness of our method in AEs pre-training tasks.

关键词： autoencoder Image reconstruction Regularization Discriminative feature

来源：评论

学校读者我要写书评

暂无评论

Mobility, Path Loss, and Composite Fading: Performance of a Conventional and of a Non-Conventional System With a Robust autoencoder

引用

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 2023年第12期72卷 16725-16730页

作者： Pereira, Pedro M. R. de Bairros, Thiago A. M. de Souza, Rausley A. A. Yacoub, Michel D. Natl Inst Telecommun BR-37540000 Santa Rita Do Sapucai Brazil Univ Estadual Campinas Dept Commun Wireless Technol Lab BR-13083852 Campinas SP Brazil

The aim of this correspondence is two-fold: (i) to assess the performance of a conventional wireless system jointly subject to short-term fading (alpha-mu model), long-term fading (Gamma model), path loss (power law decay), and mobility (random waypoint model) through analytical formulations;and (ii) to assess the performance of a non-conventional wireless system subject to the same conditions as before but with the transmission-medium-receiving chain replaced by an autoencoder using deep neural networks. The paper derives novel closed-form expressions for the probability density function, cumulative distribution function, and moment-generating function of the received signal. The performance analysis under investigation concerns outage probability, symbol error probability, asymptotic analyses, and channel capacity, all in closed-form formulations. In addition to showing that mobility has a beneficial effect on the overall mean performance, this work also illustrates that the said unconventional system may achieve as good a performance of the conventional one.

关键词： Shadowing mobility user outage probability symbol error probability channel capacity path-loss deep neural networks autoencoder random waypoint model

来源：评论

学校读者我要写书评

暂无评论

A novel asymmetrical autoencoder with a sparsifying discrete cosine Stockwell transform layer for gearbox sensor data compression

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2024年第PartB期127卷

作者： Zhu, Xin Yang, Daoguang Pan, Hongyi Karimi, Hamid Reza Ozevin, Didem Cetin, Ahmet Enis Univ Illinois Dept Elect & Comp Engn Chicago IL 60607 USA Univ Illinois Civil Mat & Environm Engn Chicago IL USA Politecn Milan Dept Mech Engn Milan Italy

The lack of an efficient compression model remains a challenge for the wireless transmission of gearbox data in non-contact gear fault diagnosis problems. In this paper, we present a signal-adaptive asymmetrical autoencoder with a transform domain layer to compress sensor signals. First, a new discrete cosine Stockwell transform (DCST) layer is introduced to replace linear layers in a multi-layer autoencoder. A trainable filter is implemented in the DCST domain by utilizing the multiplication property of the convolution. A trainable hard-thresholding layer is applied to reduce redundant data in the DCST layer to make the feature map sparse. In comparison to the linear layer, the DCST layer reduces the number of trainable parameters and improves the accuracy of data reconstruction. Second, training the autoencoder with a sparsifying DCST layer only requires a small number of datasets. The proposed method is superior to other autoencoder-based methods on the University of Connecticut (UoC) and Southeast University (SEU) gearbox datasets, as the average quality score is improved by 2.00% at the lowest and 32.35% at the highest with a limited number of training samples.

关键词： Gearbox sensor data compression autoencoder Discrete cosine stockwell transform Transform domain layer Limited samples

来源：评论

学校读者我要写书评

暂无评论

Target-Embedding autoencoder With Knowledge Distillation for Multi-Label Classification

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2024年第3期8卷 2506-2517页

作者： Ma, Ying Zou, Xiaoyan Pan, Qizheng Yan, Ming Li, Guoqi Harbin Inst Technol Fac Comp Harbin 150001 Peoples R China Xiamen Univ Technol Fac Comp & Informat Engn Xiamen 361024 Peoples R China Xiamen Univ Technol Xiamen 361021 Peoples R China Singapore Sci & Technol Bur Inst High Performance Comp Singapore Singapore Chinese Acad Sci Inst Automat Beijing 100190 Peoples R China

In the task of multi-label classification, it is a key challenge to determine the correlation between labels. One solution to this is the Target Embedding autoencoder (TEA), but most TEA-based frameworks have numerous parameters, large models, and high complexity, which makes it difficult to deal with the problem of large-scale learning. To address this issue, we provide a Target Embedding autoencoder framework based on Knowledge Distillation (KD-TEA) that compresses a Teacher model with large parameters into a small Student model through knowledge distillation. Specifically, KD-TEA transfers the dark knowledge learned from the Teacher model to the Student model. The dark knowledge can provide effective regularization to alleviate the over-fitting problem in the training process, thereby enhancing the generalization ability of the Student model, and better completing the multi-label task. In order to make the Student model learn the knowledge of the Teacher model directly, we improve the distillation loss: KD-TEA uses MSE loss instead of KL divergence loss to improve the performance of the model in multi-label tasks. Experiments on multiple datasets show that our KD-TEA framework is superior to the most advanced multi-label classification methods in both performance and efficiency.

关键词： Multi-label classification knowledge distillation autoencoder label embedding

来源：评论

学校读者我要写书评

暂无评论

ECG heartbeats classification with dilated convolutional autoencoder

引用

SIGNAL IMAGE AND VIDEO PROCESSING 2024年第1期18卷 417-426页

作者： Arslan, Naciye Nur Ozdemir, Durmus Temurtas, Hasan Kutahya Dumlupinar Univ Dept Software Engn TR-43100 Kutahya Turkiye Kutahya Dumlupinar Univ Dept Comp Engn TR-43100 Kutahya Turkiye

Electrocardiography is essential for the early diagnosis and treatment of heart diseases, as undiagnosed heart diseases can lead to unfortunate outcomes such as patient loss. autoencoder-based models have been used in the literature for ECG heartbeat classification. However, these models usually use the autoencoder in the feature extraction stage. The features obtained from the previous step are passed through a classifier for training. This indicates that the training procedure occurs in two phases. In this study, we performed autoencoder and classifier training simultaneously. This way, the network learned to minimize the overall loss while correctly reconstructing the input and extracting relevant features from the input data that are useful for the classification task. Such an approach has yet to be seen in the literature for ECG detection. The classification of six heartbeats (normal beat, left bundle branch block beat, right bundle branch block beat, premature ventricular contraction, atrial premature beat, and paced beat) obtained from the MIT-BIH dataset was performed using a convolutional autoencoder with an integrated classifier. The classification accuracy obtained in the test was found to be 99.99%.

关键词： autoencoder Convolution Classification ECG

来源：评论

学校读者我要写书评

暂无评论

Simultaneous Deep Clustering and Feature Selection via K-Concrete autoencoder

引用

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2024年第6期36卷 2629-2642页

作者： Doo, Woojin Kim, Heeyoung Korea Adv Inst Sci & Technol KAIST Dept Ind & Syst Engn Daejeon 34141 South Korea

Existing deep learning methods for clustering high-dimensional data perform feature selection and clustering separately, which can result in the exclusion of some important features for clustering. In this paper, we propose a method that performs deep clustering and feature selection simultaneously by inserting a concrete selector layer between the input layer and the first encoder layer of a modified autoencoder. The concrete selector layer performs feature selection, while the modified autoencoder performs clustering in the latent space by incorporating K-means loss and inter-cluster distances. The proposed method, called the K-concrete autoencoder, selects features important for clustering and uses only the selected features to learn K-means-friendly latent representations of the data. Moreover, we propose an extension of the K-concrete autoencoder to provide relative importance of each selected feature. We demonstrate the effectiveness of the proposed method using simulated and real datasets.

关键词： autoencoder deep clustering feature selection high-dimensional data K-means

来源：评论

学校读者我要写书评

暂无评论

Learning hash index based on a shallow autoencoder

引用

APPLIED INTELLIGENCE 2023年第12期53卷 14999-15010页

作者： Lin, Yuming Huang, Zhengguo Li, You Guilin Univ Elect Technol Guangxi Key Lab Trusted Software Jinji Rd Guilin 541004 Guangxi Peoples R China

The hash index plays an important role in improving query efficiency in databases. Because traditional hash algorithms cannot use the original data distribution, there is often a high collision rate in large-scale datasets. Additionally, some traditional hashing algorithms are data-dependent and cannot be accelerated in parallel. The learned index provides a new method for index design in data management systems. The key idea of the learned index is to consider the index as a model that can be learned. In this paper, we propose a learning hash algorithm based on a shallow autoencoder that can make full use of the original data characteristics and take advantage of the parallelism of matrix operations. Therefore, compared with traditional hash functions, the proposed method has a lower collision rate and higher efficiency. Finally, we verify the effectiveness of the proposed method through a series of experiments on synthetic datasets and real datasets. Experimental results show that the proposed hash algorithm has considerable advantages in reducing the collision rate and computing time while improving the retrieval efficiency.

关键词： Hash index Learned index autoencoder Machine learning

来源：评论

学校读者我要写书评

暂无评论

A new supervised multi-head self-attention autoencoder for health indicator construction and similarity-based machinery RUL prediction

引用

ADVANCED ENGINEERING INFORMATICS 2023年第1期56卷

作者： Qin, Yi Yang, Jiahong Zhou, Jianghong Pu, Huayan Mao, Yongfang Chongqing Univ State Key Lab Mech Transmiss Chongqing 400044 Peoples R China Chongqing Univ Coll Mech & Vehicle Engn Chongqing 400044 Peoples R China Chongqing Univ Coll Automat Chongqing 400044 Peoples R China

Remaining useful life (RUL) prediction plays a significant role in the prognostic and health management (PHM) of rotating machineries. A good health indicator (HI) can ensure the accuracy and reliability of RUL prediction. However, numerous existing deep learning-based HI construction approaches rely heavily on the prior knowledge, and they are difficult to capture the key information in the process of machinery degradation from raw signals, thereby affecting the performance of RUL prediction. To tackle the aforementioned problem, a new supervised multi-head self-attention autoencoder (SMSAE) is proposed for extracting the HI that effectively reflects the degraded state of rotating machinery. By embedding the multi-head self-attention (MS) module into autoencoder and imposing the constraint of power function-type labels on the hidden variable, SMSAE can directly extract the HIs from raw vibration signals. As the current HI evaluation indexes don't consider the global monotonicity and variation law of HI, two improved monotonicity and robustness indexes are designed for the better evaluation of HI. With the proposed HI, a two-stage residual life prediction framework based on similarity is developed. Extensive experiments have been performed on an actual wind turbine gearbox bearing dataset and a well-known open commercial modular aero-propulsion system simulation (C-MAPSS) dataset. The comparative results verify that the constructed SMSAE HI has better comprehensive performance than the typical HIs, and the proposed prediction method is competitive with the state-of-the-art methods.

关键词： Health indicator RUL prediction autoencoder Self-attention Rotating machinery

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：