Search

Chronic obstructive pulmonary disease (COPD) is a complex and multi-component respiratory disease. Computed tomography (CT) images can characterize lesions in COPD patients, but the image intensity and morphology of lung components have not been fully exploited. Two datasets (Dataset 1 and 2) comprising a total of 561 subjects were obtained from two centers. A multiple instance learning (MIL) method is proposed for COPD identification. First, randomly selected slices (instances) from CT scans and multi-view 2D snapshots of the 3D airway tree and lung field extracted from CT images are acquired. Then, three attention-guided MIL models (slice-CT, snapshot-airway, and snapshot-lung-field models) are trained. In these models, a deep convolution neural network (CNN) is utilized for feature extraction. Finally, the outputs of the above three MIL models are combined using logistic regression to produce the final prediction. For Dataset 1, the accuracy of the slice-CT MIL model with 20 instances was 88.1%. The backbone of VGG-16 outperformed Alexnet, Resnet18, Resnet26, and Mobilenet_v2 in feature extraction. The snapshot-airway and snapshot-lung-field MIL models achieved accuracies of 89.4% and 90.0%, respectively. After the three models were combined, the accuracy reached 95.8%. The proposed model outperformed several state-of-the-art methods and afforded an accuracy of 83.1% for the external dataset (Dataset 2). The proposed weakly supervised MIL method is feasible for COPD identification. The effective CNN module and attention-guided MIL pooling module contribute to performance enhancement. The morphology information of the airway and lung field is beneficial for identifying COPD. [ABSTRACT FROM AUTHOR]

Due to a shortage of labeled examples, few-shot image classification frequently experiences noise interference and insufficient feature extraction. In this paper, we present a two-stage framework based on the distribution propagation graph neural network (DPGN) called the multilevel distribution propagation network (MDPN). An instance-segmentation-based object localization (ISOL) module and a graph-based multilevel distribution propagation (GMDP) module are both included in the MDPN. To create a clear and full object zone, the ISOL module generates a mask that eliminates background and pseudo-object noises. The GMDP module enriches the level of features. We carried out comprehensive experiments on the few-shot dataset CUB-200-2011 to show the usefulness of MDPN. The results demonstrate that MDPN indeed outperforms DPGN in terms of few-shot image classification accuracy. Under 5-way 1-shot and 5-way 5-shot settings, the classification accuracy of MDPN exceeds the baseline by 8.17% and 1.24%, respectively. MDPN also outperforms the majority of the existing few-shot classification methods in the same setting. [ABSTRACT FROM AUTHOR]

Obtaining crowd flow distribution with recognized human intention is extremely valuable for a series of applications for metropolitan cities. Previous solutions look at spatial correlation and temporal periodicity based on historical crowd flow information to calculate future crowd flow distribution. However, these mechanisms cannot recognize the intention behind crowd flow. We address this problem by leveraging a key insight – people's intention behind their movement is highly correlated with the point-of-interest (POI) distribution of the corresponding regions and adjacent regions. Therefore, we propose DeepFlowGen to model the complicated relationship between crowd flow, POI, check-ins, and time to generate intention-aware crowd flow. Specifically, we solve the conflict between dynamic crowd flow and static POI distribution by fusing the information in both time and POI domains. Besides, we employ a sequence of residual blocks in DeepFlowGen to address the challenges of modeling the diverse temporal rhythms and heterogeneous influence of POI. Furthermore, we examine the generated intention-aware crowd flow from two aspects to substantiate the reasonability of DeepFlowGen. Extensive experiments demonstrate that our model outperforms the state-of-the-art solutions by at most 30 percent in terms of NRMSE of total crowd flow. Moreover, the correlation between the generated intention-aware crowd flow and the check-in distribution across different categories of POIs is as high as 0.90 and 0.80 in Beijing and Shanghai. Combined with extensive case studies, we demonstrate the strong ability of our model in generating intention-aware crowd flow. [ABSTRACT FROM AUTHOR]

Recently, most convolutional neural network-based methods use convolutional kernels of fixed size to extract features, which ignore the inherent spatial structure information of ground objects and lose spatial details. In addition, rough first-order statistics is not enough to capture subtle differences between different categories and extract nonlocal context information. To address these issues, a hyperspectral image (HSI) classification method based on multiscale cross-branch response and second-order channel attention (MCRSCA) is proposed in this article. First, a multiscale cross-branch response (MCBR) module is proposed, which uses convolution kernels of different sizes for feature extraction. It adds and concatenates the features of different scales, respectively, to obtain rich and complementary spatial context information. Then, element multiplication and element addition are performed on the fused multiscale features to promote the propagation of the multiscale information and enhance the nonlinear expression ability. Next, the second-order channel attention (SOCA) module is designed to interact the channel information through the feature covariance matrix to obtain the long-term dependence between channels. This module pays more attention to the significant channels and suppresses the redundant channels. Finally, the residual connection is used to embed MCBR and SOCA into the residual block to improve the gradient back propagation and accelerate the training process. Experiments on four commonly used HSI benchmark datasets show that the results of MCRSCA are competitive compared with other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Deep learning methods have gained rapid development in hyperspectral pansharpening (HP) due to powerful spatial–spectral feature extraction ability. However, most of these methods are optimized using a single reconstruction objective. It is difficult for these methods to find a balance between spectral preservation and spatial preservation. Furthermore, these methods adopt interpolation or convolution to upsample the hyperspectral images (HSIs), which tends to cause noticeable spectral distortion. To conquer these issues, a novel multiobjective guided divide-and-conquer network (MO-DCN) is proposed for HP. It consists of a deconvolution long short-term memories (LSTMs) network (DLSTM) and a divide-and-conquer network (DCN). DLSTM leverages bi-direction learning to upsample HSIs by considering 3-D spatiotemporal dependencies. Then, DCN designs a two-branch architecture to reconstruct spatial and spectral information from upsampled HSIs and panchromatic images (PANIs), respectively, where the spatial branch designs an attention-in-attention module (AIAM) to emphasize complementary attention in a coarse-to-fine way. Finally, co-improvement of spatial and spectral information is formulated as an Epsilon-constraint-based multiobjective optimization. The Epsilon constraint method transforms one objective into a constraint and regards it as a penalty bound to make an excellent tradeoff between different objectives. Experimental results demonstrated that the proposed method markedly improves pansharpening performance in both the spatial and spectral domains and has superior fusion performance than state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Graph convolution networks (GCNs) have been applied in a variety of fields due to their powerful ability in processing graph-like data. However, the massive number of hyperspectral pixels makes it challenging to define general graph structures on hyperspectral images (HSIs). On the other hand, convolutional neural networks (CNNs) take in regular image regions with fixed square size, and have demonstrated impressive accuracy while being efficient in computation. Inspired by the classification framework of CNNs, we develop a GCN-based model that generates effective local spectral–spatial features for HSI classification. Specifically, graph convolutions are performed separately on every local region, which significantly limits the graph’s size. While graph convolution extracts features of every pixel, it does not reduce the number of them. To fuse suitable representations for the classification task, we develop a graph pooling operation to preserve classification-specific features and reduce redundant pixels. Based on local regions of HSIs, pooling in the graph domain is equivalent to spatial pooling in the spatial domain. The proposed method is thus named the spatial pooling graph convolutional network (SPGCN). Experimental results on several typical datasets demonstrated that the proposed SPGCN provides competitive results compared with other state-of-the-art CNN-based methods. [ABSTRACT FROM AUTHOR]

Online services are playing critical roles in almost all aspects of users’ life. Users usually have multiple online identities (IDs) in different online services. In order to fuse the separated user data in multiple services for better business intelligence, it is critical for service providers to link online IDs belonging to the same user. On the other hand, the popularity of mobile networks and GPS-equipped smart devices have provided a generic way to link IDs, i.e., utilizing the mobility traces of IDs. However, linking IDs based on their mobility traces has been a challenging problem due to the highly heterogeneous, incomplete and noisy mobility data across services. In this paper, we propose DPLink, an end-to-end deep learning based framework, to complete the user identity linkage task for heterogeneous mobility data collected from different services with different properties. DPLink is made up by a feature extractor including a location encoder and a trajectory encoder to extract representative features from trajectory and a comparator to compare and decide whether to link two trajectories as the same user. Particularly, we propose a pre-training strategy with a simple task to train the DPLink model to overcome the training difficulties introduced by the highly heterogeneous nature of different source mobility data. Besides, we introduce a multi-modal embedding network and a co-attention mechanism in DPLink to deal with the low-quality problem of mobility data. By conducting extensive experiments on two real-life ground-truth mobility datasets with eight baselines, we demonstrate that DPLink outperforms the state-of-the-art solutions by more than 15 percent in terms of hit-precision. Moreover, it is expandable to add external geographical context data and works stably with heterogeneous noisy mobility traces. [ABSTRACT FROM AUTHOR]

For hyperspectral imagery (HSI) classification, most of the studies focus on how to improve the classification accuracy, while the influence of sampling strategy for classification performance attracts little attention. For now, random sampling (RS) is the most adopted strategy. That is, for a hyperspectral image, a certain number of labeled samples are randomly selected as the training set, and the remaining labeled samples are taken as the test set. However, the RS strategy will produce over optimistic results when used for performance evaluation because of the overlap between training set and test set. Though spectral-spatial classification methods benefit most from the RS strategy, the pixel-wise classification methods can also benefit from it because of the high spectral correlation between training and test samples. However, in practical applications, the RS strategy is not feasible. Because the training and test samples are often collected from different locations. In this situation, the correlation between training and test samples will decrease dramatically and the performance of HSI classification methods will be affected. In this article, a nonoverlapped sampling method is adopted to reduce the correlation between training and test samples and different classic classification methods are evaluated. Experimental results show that the classification performance of all methods drops a lot when nonoverlapped sampling strategy is adopted. After the analysis of some important factors for HSI classification, we also propose a cotraining-based classification method to relief the influence of sampling strategy and obtains much better performance compared with those classic spectral-spatial classification methods. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) classification is one of the most important tasks in hyperspectral data analysis. Convolutional neural networks (CNN) have been introduced to HSI classification and achieved good performance. In this article, an effective and efficient CNN-based spectral partitioning residual network (SPRN) is proposed for HSI classification. The SPRN splits the input spectral bands into several nonoverlapping continuous subbands and uses cascaded parallel improved residual blocks to extract spectral–spatial features from these subbands, respectively. Finally, the features are fused and fed into a classifier. By equivalently using grouped convolutions, the spectral partition and feature extraction are embedded into an end-to-end network. Experimental results show that the proposed SPRN achieves state-of-the-art performance, meanwhile, with relatively fewer parameters and computational costs. Usually, the CNN takes a patch that contains continuous spatial information as the input and results in a class label of the center pixel. The large size of the input patch includes more spatial information, whereas also introduces interfering pixels that may lead to a degradation of classification accuracies. For that reason, we propose a novel spatial attention module named homogeneous pixel detection module (HPDM). The module alleviates the degradation of performance as the input patch size increases by capturing the homogeneous pixels in the input patch. The module can be integrated into any CNN-based HSI classification framework. [ABSTRACT FROM AUTHOR]

Band selection has been widely utilized in hyperspectral image (HSI) classification to reduce the dimensionality of HSIs. Recently, deep-learning-based band selection has become of great interest. However, existing deep-learning-based methods usually implement band selection and classification in isolation, or evaluate selected spectral bands by training the deep network repeatedly, which may lead to the loss of discriminative bands and increased computational cost. In this article, a novel convolutional neural network (CNN) based on bandwise-independent convolution and hard thresholding (BHCNN) is proposed, which combines band selection, feature extraction, and classification into an end-to-end trainable network. In BHCNN, a band selection layer is constructed by designing bandwise $1\times 1$ convolutions, which perform for each spectral band of input HSIs independently. Then, hard thresholding is utilized to constrain the weights of convolution kernels with unselected spectral bands to zero. In this case, these weights are difficult to update. To optimize these weights, the straight-through estimator (STE) is devised by approximating the gradient. Furthermore, a novel coarse-to-fine loss calculated by full and selected spectral bands is defined to improve the interpretability of STE. In the subsequent layers of BHCNN, multiscale 3-D dilated convolutions are constructed to extract joint spatial–spectral features from HSIs with selected spectral bands. The experimental results on several HSI datasets demonstrate that the proposed method uses selected spectral bands to achieve more encouraging classification performance than current state-of-the-art band selection methods. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs) have demo- nstrated outstanding performance on image classification. To classify the hyperspectral images (HSIs), existing CNN-based approaches commonly adopt the architecture using single or several fixed spatial windows as inputs. This kind of architecture may lose contextual information or incorporate heterogeneous information due to the neglect of various land-cover distributions in HSIs. To deal with this problem, a novel attention multibranch CNN method based on adaptive region search (RS-AMCNN) is proposed for HSI classification. In RS-AMCNN, sizes and locations of spatial windows are searched in the nonlocal candidate region adaptively according to sample-specific distribution. These flexible spatial windows are input into several branches of RS-AMCNN. In each branch, convolutional long short-term memories (ConvLSTMs) are merged into CNN from shallow to deep layers, which not only extracts joint spatial–spectral features, but also exploits complementary information among different layers. Then, a branch attention mechanism is devised to emphasize more discriminative branches and suppress less useful ones. It forces RS-AMCNN to extract multiscale and multicontextual attention features for classification. Finally, RS-AMCNN is optimized end-to-end by combining the losses from the ramose classifiers of different branches and the main classifier. Experiments carried on several benchmark HSI data sets demonstrate that RS-AMCNN provides promising classification performance, especially in edge preservation and region uniformity. [ABSTRACT FROM AUTHOR]

Pan sharpening is used to fuse a low-resolution multispectral (MS) image and a high-resolution panchromatic (PAN) image to obtain a high-resolution MS image. This article proposes PSMD-Net, an end-to-end pan-sharpening method based on a multi-scale dense network. A shallow feature extraction layer (SFEL) extracts the shallow features from the original images, and these are used as an input to a global dense feature fusion (GDFF) network to learn the global features for image reconstruction. A multiscale dense block (MDB) is designed to fully extract the spatial and spectral information from the shallow features in the GDFF network. In the proposed network, multiple MDBs are stacked to extract rich, multi-scale dense hierarchical features, and a global dense connection (GDC) is designed to allow direct connections from the state of the current MDB to all subsequent MDBs to extract more advanced features. The extracted hierarchical features are sent to the global feature fusion layer (GFFL) to adaptively learn the global features for image reconstruction. Finally, global residual learning (GRL) is adopted to force the network to pay more attention to the changing part of the image. We perform experiments on simulated and real data from WorldView-2 and WorldView-3 satellites. Visual and quantitative assessment results demonstrate that PSMD-Net yields higher-resolution fusion images than the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Generative adversarial networks (GANs) are famous for generating samples by training a generator and a discriminator via an adversarial procedure. For hyperspectral image classification, the collection of samples is always difficult. However, directly applying GAN to hyperspectral image classification exists two problems. One is that the generated samples lack discriminative information. Meanwhile, the discriminator has no discriminative ability for multiclassification. Another is that spatial and spectral information requires to be considered in hyperspectral image classification simultaneously. To address these problems, a novel multiclass spatial–spectral GAN (MSGAN) method is proposed. In MSGAN, two generators are devised to generate the samples containing spatial and spectral information, respectively, and the discriminator is devised to extract joint spatial–spectral features and output multiclass probabilities. Moreover, novel adversarial objectives for multiclass are defined. The discriminator is devised to predict training samples belonging to true classes and generated samples belonging to all the classes with the same probability. The generators are devised to make the discriminator mistake. By adversarial learning between the discriminator and generators, the classification performance of the discriminator is promoted with the assistance of discriminative generated samples. Experimental results on hyperspectral images demonstrate that the proposed method achieves encouraging classification performance compared with several state-of-the-art methods, especially with the limited training samples. [ABSTRACT FROM AUTHOR]

The extraction of joint spatial–spectral features has been proved to improve the classification performance of hyperspectral images (HSIs). Recently, utilizing convolutional neural networks (CNNs) to learn joint spatial–spectral features has become of great interest. However, the existing CNN models ignore complementary spatial–spectral information among the shallow and deep layers. Moreover, insufficient training samples in HSIs afflict these CNN models with overfitting problem. In order to address these problems, a novel CNN method for HSI classification is proposed. It considers multilayer spatial–spectral feature fusion and sample augmentation with local and nonlocal constraints, which is abbreviated as MSLN-CNN. In MSLN-CNN, a triple-architecture CNN is constructed to extract spatial–spectral features by cascading spectral features to dual-scale spatial features from shallow to deep layers. Then, multilayer spatial–spectral features are fused to learn complementary information among the shallow layers with detailed information and the deep layers with semantic information. Finally, the multilayer spatial–spectral feature fusion and classification are integrated into a unified network, and MSLN-CNN can be optimized in the end-to-end way. To alleviate the small sample size problem, the unlabeled samples having high confidences on local spatial constraint and nonlocal spectral constraint are selected and prelabeled. The nonlocal spectral constraint considers the structure information with spectrally similar samples in the nonlocal searching, while the local spatial constraint utilizes the contextual information with spatially adjacent samples. Experimental results on several hyperspectral datasets demonstrate that the proposed method achieves more encouraging classification performance than the current state-of-the-art classification methods, especially with the limited training samples. [ABSTRACT FROM AUTHOR]

The abundant spectral information of hyperspectral imagery makes it suitable for the classification of land cover types. However, the high dimensionality also brings some negative effects for the classification tasks. Dynamic classifier selection, in which the base classifiers are selected according to each new sample to be classified, can select the best classifier for each query sample. In this paper, a semi-supervised wrapper band selection method—the band selection based on dynamic classifier selection—is introduced to select the most discriminating bands. In the proposed method, band selection is conducted based on the selection of base classifier. Specifically, the support vector machine classification map is filtered to provide a high-quality reference, and K-nearest neighbors method is used to define the local region, finally, the band with the best classification performance is selected. Three widely used real hyperspectral datasets are used to illustrate the effectiveness of the proposed method, experimental results show that the proposed method obtains state-of-the-art performance. [ABSTRACT FROM AUTHOR]

The past 10 years have witnessed the rapid growth of global mobile cellular traffic demands due to the popularity of mobile devices. While accurate traffic prediction becomes extremely important for stable and high-quality Internet service, the performance of existing methods is still poor due to three challenges: complicated temporal variations including burstiness and long periods, multi-variant impact factors such as the point of interest and day of the week, and potential spatial dependencies introduced by the movement of population. While existing traditional methods fail in characterizing these features, especially the latter two, deep learning models with powerful representation ability give us a chance to consider these from a new perspective. In this article, we propose Deep Traffic Predictor (DeepTP), a deep-learning-based end-toend model, which forecasts traffic demands from spatial-dependent and long-period cellular traffic. DeepTP consists of two components: a general feature extractor for modeling spatial dependencies and encoding the external information, and a sequential module for modeling complicated temporal variations. In the general feature extractor, we introduce a correlation selection mechanism for a spatial modeling and embedding mechanism to encode external information. Moreover, we apply a seq2seq model with attention mechanism to build the sequential model. Extensive experiments based on large-scale mobile cellular traffic data demonstrate that our model outperforms the state-of-the-art traffic prediction models by more than 12.31 percent. [ABSTRACT FROM AUTHOR]

Hyperspectral images usually have high-dimensional and abundant spectral information for land-cover types classification. Research shows that multiple kinds of features would be helpful to the classification task. In this paper, a new feature fusion framework—deep multiple feature fusion (DMFF)—is proposed for the classification of the hyperspectral image. First, several different features are extracted for each pixel. Then, these features are fed to a deep random forest classifier. With a multiple-layer structure, the outputs of preceding layers will be used as the inputs of the subsequent layers. After the final layer, the classification probability will be computed. By introducing the information of neighboring pixels, the spectral–spatial information is combined effectively. Besides, the structure of the DMFF is easy to expand. Experimental results based on two widely used hyperspectral datasets (Indian pines image and Pavia University image) demonstrate that the proposed method can achieve a satisfactory classification performance compared with other multiple feature fusion methods. [ABSTRACT FROM AUTHOR]

Stacked autoencoder (SAE) provides excellent performance for image processing under sufficient training samples. However, the collection of training samples is difficult in hyperspectral images. Insufficient training samples easily make SAE overfit and limit the application of SAE to hypersepctral images. To address this problem, a novel marginal SAE with adaptively-spatial regularization (ARMSAE) is proposed for hyperspectral image classification. First, a superpixel segmentation method is used to divide the image into many homogenous regions. Then, at the pretraining stage, an adaptively-shaped spatial regularization is introduced to extract contextual information of samples in the homogenous regions. It sufficiently utilizes unlabeled adjacent samples to alleviate the lack of training samples. At the fine-tuning stage, the marginal samples based on geometrical property are selected to tune the ARMSAE network. The fine-tuning exploits margin strategy to alleviate the inaccurate statistical estimation caused by insufficient training samples. Finally, the label of each test sample is determined by all the samples locating in the same homogenous region. Experimental results on hyperspectral images demonstrate the proposed method provides encouraging classification performance compared with several related state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Hyperspectral pansharpening (HP) can fuse different kinds of remote sensing images for the downstream classification or detection task. Deep learning methods have made rapid development in HP. However, most deep learning-based HP methods design the network architecture without considering the specific downstream task. When the pansharpening image is transferred to the downstream tasks, it may lead to poor performance on the downstream tasks. To solve this problem, a novel classification-oriented multi-task network (CMNet) is proposed for HP. CMNet combines the classification and HP networks in a multi-task learning way. In the classification network, class-specific and spatial-difference contrastive constraints are defined to enforce the extraction module to extract discriminative features for classification. In the HP network, a fusion module with an identical structure to the extraction module is constructed to extract fused features for reconstruction. Then, multi-gradient learning is designed to optimize the HP network under the guidance of the classification network. It adopts gradient mapping to consider the conflicting or collaborative situations between these different tasks. In this way, the fusion module in the HP network learns discriminative feature extraction ability from the classification network, which makes the HP network realize classification-oriented HP. Experimental results demonstrate that the proposed method markedly improves the pansharpening performance and the pansharpening is also able to elevate the classification performance. [ABSTRACT FROM AUTHOR]

Recently, graph-based semi-supervised classification (SSC) has attracted considerable attentions as it could enhance classification accuracy by utilizing only a few labeled samples and large numbers of unlabeled samples via graphs. However, the construction of graphs is time consuming especially for large-scale polarimetric synthetic aperture radar (PolSAR) data. Moreover, speckle noise in images remarkably degrades the accuracy of the constructed graph. To address these two issues, this paper proposes a novel spatial-anchor graph for large-scale PolSAR terrain classification. First, the PolSAR image is segmented to obtain homogeneous regions. The features of each pixel are weighted by that of the surrounding pixels from the homogeneous regions to reduce the influence of the speckle noise. Second, Wishart distance-based clustering is performed on the weighted features, and the cluster centers are computed and serve as initial anchors. Then, the label of each pixel is predicted by the label of its nearest anchors on the spatial-anchor graph which is constructed through solving an optimization problem. Experimental results on synthesized PolSAR data and real ones from different approaches show that the proposed method reduces the computational complexity to a linear time, and the graph combined with the spatial information suppresses the speckle noise and enhances the classification accuracy in comparison with state-of-the-art graph-based SSCs when only a small number of labeled samples are available. [ABSTRACT FROM PUBLISHER]

Band selection is a research hotspot in hyperspectral image processing. The continuity of the spectral bands causes the adjacent bands to be highly correlated, and correlation among long-range bands is possible with hundreds of spectral bands. Most existing deep learning methods fail to make full use of the inter-band correlation for band selection. In this paper, a novel dual-graph convolutional network based on band attention and a sparse constraint is proposed for band selection. The network consists of two branches. In the attention branch, band-based dual graphs are constructed to encode the contextual correlation of adjacent bands and the structural correlation of long-range bands into non-Euclidean space. Subsequently, the graph convolution-based band attention mechanism is devised to aggregate the band information in the band-based dual graphs and to generate the attention map for all bands. The band attention map is sparsely constrained and embedded as a mask into the trunk branch. In the trunk branch, sample-based dual graphs are constructed to represent the topological information of the samples in the spectral and spatial domains. Furthermore, a dense graph convolutional network is designed to extract and fuse the spatial–spectral and topological features from the shallow to deep layers for classification. A soft-shifting optimization strategy is implemented by defining a new loss from full bands and selected bands to solve the optimization problem caused by the sparse constraint. In this manner, band selection, feature extraction, and classification can be combined into an end-to-end trainable network. The experimental results on representative hyperspectral image datasets demonstrate the superiority of the proposed method over current state-of-the-art band selection methods. • An improved graph convolutional network is proposed for band selection. • Band-based and sample-based graphs are built with spatial and spectral information. • Design soft-shifting optimization strategy to overcome the optimization problem. • Mini batches of labeled and unlabeled samples are used for graph construction. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) image classification involves two crucial issues: suitable feature representation technique and effective pattern classification methodology. Here, we concentrate on the first issue. By exploiting a famous image feature processing strategy, Bag-of-Visual-Words (BOV) in image semantic analysis and the artificial immune systems (AIS)'s abilities of learning and adaptability to solve complicated problems, we present a novel and effective image representation method for SAR image classification. In BOV, an effective fused feature sets for local feature representation are first formulated, which are viewed as the low-level features in it. After that, clonal selection algorithm (CSA) in AIS is introduced to optimize the prediction error of k-fold cross-validation for getting more suitable visual words from the low-level features. Finally, the BOV features are represented by the learned visual words for subsequent pattern classification. Compared with the other four algorithms, the proposed algorithm obtains more satisfactory and cogent classification experimental results. [ABSTRACT FROM PUBLISHER]

• A novel end-to-end heterogeneous graph-based deep learning model. • Capturing the internal temporal pattern of single-dimensional time series and the rich spatial relations among variables. • Capturing both the static and dynamic relations of multivariate time series (MTS). • Combining graph neural networks with MTS to take full advantage of complex relations among variables of MTS. Multivariate time series forecasting, which analyzes historical time series to predict future trends, can effectively help decision-making. Complex relations among variables in MTS, including static, dynamic, predictable, and latent relations, have made it possible to mining more features of MTS. Modeling complex relations are not only essential in characterizing latent dependency as well as modeling temporal dependence, but also brings great challenges in the MTS forecasting task. However, existing methods mainly focus on modeling certain relations among MTS variables. In this paper, we propose a novel end-to-end deep learning model, termed M ultivariate T ime Series Forecasting via Het erogeneous G raph N eural N etworks (MTHetGNN). To characterize complex relations among variables, a relation embedding module is designed in MTHetGNN, where each variable is regarded as a graph node, and each type of edge represents a specific static or dynamic relationship. Meanwhile, a temporal embedding module is introduced for time series features extraction, where involving convolutional neural network (CNN) filters with different perception scales. Finally, a heterogeneous graph embedding module is adopted to handle the complex structural information generated by the two modules. Three benchmark datasets from the real world are used to evaluate the proposed MTHetGNN. The comprehensive experiments show that MTHetGNN achieves state-of-the-art results in the MTS forecasting task. [ABSTRACT FROM AUTHOR]