The problem of generalized zero-shot learning deals with the classification of test examples for which training data may or may not be available. Existing baseline algorithms connect the seen and unseen set of categor...
详细信息
ISBN:
(纸本)9781538662496
The problem of generalized zero-shot learning deals with the classification of test examples for which training data may or may not be available. Existing baseline algorithms connect the seen and unseen set of categories by learning functions to project the image data into the attribute space or vice versa. However, since the classification framework is trained only on the seen set of categories, the recognition performance is typically biased and algorithms have great difficulty in recognizing novel classes. In this work, we investigate the usefulness of a novelty detector to recognize a given data as coming from the seen or novel set. The proposed novelty detector is based on an autoencoder network structure with reconstruction and triplet cosine embedding losses which can be effectively trained using only the seen data and its categories. Experiments over a variety of benchmark datasets and zero-shot algorithms show the efficacy of the proposed approach.
In the past decade,recommender systems have been widely used to provide users with personalized products and ***,most traditional recommender systems are still facing a challenge in dealing with the huge volume,comple...
详细信息
In the past decade,recommender systems have been widely used to provide users with personalized products and ***,most traditional recommender systems are still facing a challenge in dealing with the huge volume,complexity,and dynamics of *** tackle this challenge,many studies have been conducted to improve recommender system by integrating deep learning *** an unsupervised deep learning method,autoencoder has been widely used for its excellent performance in data dimensionality reduction,feature extraction,and data ***,recent researches have shown the high efficiency of autoencoder in information retrieval and recommendation *** autoencoder on recommender systems would improve the quality of recommendations due to its better understanding of users,demands and characteristics of *** paper reviews the recent researches on autoencoder-based recommender *** differences between autoencoder-based recommender systems and traditional recommender systems are presented in this *** last,some potential research directions of autoencoder-based recommender systems are discussed.
Accurate outdoor illumination estimation is not easy due to extremely complicated sky appearance and the mutual interference of the sun and sky. In view of these challenges, we present a new deep approach for the esti...
详细信息
Accurate outdoor illumination estimation is not easy due to extremely complicated sky appearance and the mutual interference of the sun and sky. In view of these challenges, we present a new deep approach for the estimation of all-weather outdoor illumination. The key to our approach is a novel dual attention autoencoder(DAA) with two independent branches to compress the sun and sky lighting information from an input HDR panorama, respectively. This enables more accurate lighting estimation as evidenced by our experiments since the mutual interference between the sun and sky can be precluded effectively. In DAA,we design the adaptive feature pyramid and the attention module to promote its accuracy in *** further develop a sun-sky predictor, a masked network, to learn the sun and sky lighting conditions from an Fo V-limited image. Comprehensive qualitative and quantitative experiments verify the effectiveness of our proposed approach and show its superiority over the state of the arts.
Sensor networks play an essential role in today's air quality monitoring platforms. Nevertheless, sensors often malfunction, leading to data anomalies. In this paper, an unsupervised pattern-based attention recurr...
详细信息
Sensor networks play an essential role in today's air quality monitoring platforms. Nevertheless, sensors often malfunction, leading to data anomalies. In this paper, an unsupervised pattern-based attention recurrent autoencoder for anomaly detection (PARAAD) is proposed to detect and locate anomalies in a network of air quality sensors. The novelty of the proposal lies in the use of temporal patterns, i.e., blocks of data, instead of point values. By looking at temporal patterns and through an attention mechanism, the architecture captures data dependencies in the feature space and latent space, enhancing the model's ability to focus on the most relevant parts. Its performance is evaluated with two categories of anomalies, bias fault and drift anomalies, and compared with baseline models such as a feed-forward autoencoder and a transformer architecture, as well as with models not based on temporal patterns. The results show that PARAAD achieves anomalous sensor detection and localization rates higher than 80%, outperforming existing baseline models in air quality sensor networks for both bias and drift faults.
Effective health indicator (HI) construction can help equipment managers detect the abnormal state of rotating machinery quickly. However, although the current deep learning-based HI construction methods have good lif...
详细信息
Effective health indicator (HI) construction can help equipment managers detect the abnormal state of rotating machinery quickly. However, although the current deep learning-based HI construction methods have good life prediction value, most of them lose the ability to detect device anomalies and little work has been done on model interpretability. Therefore, an interpretable HI construction method based on semi-supervised autoencoder (AE) latent space variance maximization (SSALSVM) was proposed to monitor the health status of bearings. In order to fully excavate degradation features inside the device and make the model focus on the encoding process, a deep convolutional neural network (DCNN) is used as the encoding layer, while only a layer of fully-connected layer is used as the decoding layer. In addition, to enable the latent space to capture the device early degradation point (EDP) successfully, an auxiliary layer is added to the output of the encoder layer. Simultaneously, for improving the sensitivity of the indicator to capture equipment abnormal state and highlight the difference between equipment health state and degradation state, the constraint of variance maximization is added into the latent space. The model optimizing process was presented by observing the projected variance of the test set in latent space of each epoch model. The validity of the proposed HI was verified by comparison experiments on two datasets.
Hyperspectral (HS) data comprises of continuous spectral responses of hundreds of narrow spectral bands with very fine spectral resolution or bandwidth, which offer feature identification and classification with high ...
详细信息
Hyperspectral (HS) data comprises of continuous spectral responses of hundreds of narrow spectral bands with very fine spectral resolution or bandwidth, which offer feature identification and classification with high accuracy. In the present study, Mutual Information (MI) based Segmented Stacked autoencoder (S-SAE) approach for spectral-spatial classification of the HS data is proposed to reduce the complexity and computational time compared to Stacked autoencoder (SAE) based feature extraction. A non-parametric dependency measure (MI) based spectral segmentation is proposed instead of linear and parametric dependency measure to take care of both linear and nonlinear inter-band dependency for spectral segmentation of the HS bands. Then morphological profiles are created corresponding to segmented spectral features to assimilate the spatial information in the spectral-spatial classification approach. Two non-parametric classifiers, Support Vector Machine (SVM) with Gaussian kernel and Random Forest (RF) are used for classification of the three most popularly used HS datasets. Results of the numerical experiments carried out in this study have shown that SVM with a Gaussian kernel is providing better results for the Pavia University and Botswana datasets whereas RF is performing better for Indian Pines dataset. The experiments performed with the proposed methodology provide encouraging results compared to numerous existing approaches. (C) 2018 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.
A key challenge in building machine learning models for time series prediction is the incompleteness of the datasets. Missing data can arise for a variety of reasons, including sens...
详细信息
A key challenge in building machine learning models for time series prediction is the incompleteness of the datasets. Missing data can arise for a variety of reasons, including sensor failure and network outages, resulting in datasets that can be missing significant periods of measurements. Models built using these datasets can therefore be biased. Although various methods have been proposed to handle missing data in many application areas, more air quality missing data prediction requires additional investigation. This study proposes an autoencoder model with spatiotemporal considerations to estimate missing values in air quality data. The model consists of one-dimensional convolution layers, making it flexible to cover spatial and temporal behaviours of air contaminants. This model exploits data from nearby stations to enhance predictions at the target station with missing data. This method does not require additional external features, such as weather and climate data. The results show that the proposed method effectively imputes missing data for discontinuous and long-interval interrupted datasets. Compared to univariate imputation techniques (most frequent, median and mean imputations), our model achieves up to 65% RMSE improvement and 20–40% against multivariate imputation techniques (decision tree, extra-trees, k-nearest neighbours and Bayesian ridge regressors). Imputation performance degrades when neighbouring stations are negatively correlated or weakly correlated.
Unsupervised feature selection is one of the efficient approaches to reduce the dimension of unlabeled high-dimensional data. We present a novel adaptive autoencoder with redundancy control (AARC) as an unsupervised f...
详细信息
Unsupervised feature selection is one of the efficient approaches to reduce the dimension of unlabeled high-dimensional data. We present a novel adaptive autoencoder with redundancy control (AARC) as an unsupervised feature selector. By adding two Group Lasso penalties to the objective function, AARC integrates unsupervised feature selection and determination of a compact network structure into a single framework. Besides, a penalty based on a measure of dependency between features (such as Pearson correlation, mutual information) is added to the objective function for controlling the level of redundancy in the selected features. To realize the desired effects of different regularizers in different phases of the training, we introduce adaptive parameters which change with iterations. In addition, a smoothing function is utilized to approximate the three penalties since they are not differentiable at the origin. An ablation study is carried out to validate the capabilities of redundancy control and structure optimization of AARC. Subsequently, comparisons with nine state-of-the-art methods illustrate the efficiency of AARC for unsupervised feature selection. (c) 2022 Elsevier Ltd. All rights reserved.
Due to its high accuracy and ease of calculation,synchrophasor-based linear state estimation(LSE)has attracted a lot of attention in the last decade and has formed the cornerstone of many wide area monitor system(WAMS...
详细信息
Due to its high accuracy and ease of calculation,synchrophasor-based linear state estimation(LSE)has attracted a lot of attention in the last decade and has formed the cornerstone of many wide area monitor system(WAMS)***,an increasing number of data quality concerns have been reported,among which bad data can significantly undermine the performance of LSE and many other WAMS applications it *** data filtering can be difficult in practice due to a variety of issues such as limited processing time,non-uniform and changing patterns,and *** pre-process phasor measurement unit(PMU)measurements for LSE,we propose an improved denoising autoencoder(DA)-aided bad data filtering strategy in this *** data is first identified by the classifier module of the proposed DA and then recovered by the autoencoder *** characteristics distinguish the proposed methodology:1)The approach is lightweight and can be implemented at individual PMU level to achieve maximum parallelism and high efficiency,making it suited for real-time processing;2)the system not only identifies bad data but also recovers it,especially for critical *** use numerical experiments employing both simulated and real-world phasor data to validate and illustrate the effectiveness of the proposed method.
Clustering plays a crucial role in the field of data mining, where deep non-negative matrix factorization (NMF) has attracted significant attention due to its effective data representation. However, deep matrix factor...
详细信息
Clustering plays a crucial role in the field of data mining, where deep non-negative matrix factorization (NMF) has attracted significant attention due to its effective data representation. However, deep matrix factorization based on autoencoder is typically constructed using multi-layer matrix factorization, which ignores nonlinear mapping and lacks learning rate to guide the update. To address these issues, this paper proposes an autoencoder-like deep NMF representation learning (ADNRL) algorithm for clustering. First, according to the principle of autoencoder, construct the objective function based on NMF. Then, decouple the elements in the matrix and apply the nonlinear activation function to enforce non-negative constraints on the elements. Subsequently, the gradient values corresponding to the elements update guided by the learning rate are transformed into the weight values. This weight values are combined with the activation function to construct the ADNRL deep network, and the objective function is minimized through the learning of the network. Finally, extensive experiments are conducted on eight datasets, and the results demonstrate the superior performance of ADNRL.
暂无评论