Non-invasive neural activity decoding of speech perception (e.g., MEG) is of great significance for both artificial intelligence and cognitive neuroscience. Multimodal contrastive learning (MCL) aligns trainable brain activity representations with pre-trained model representations, reducing reliance on scarce MEG data and yielding promising results. However, existing methods still face limitations in modeling cross-modal consistency, primarily due to inherent distributional discrepancies, semantic misalignment, and noise between modalities such as audio signals and neural data, which lead to biased cross-modal mappings. To address these challenges, we propose an Aligned Audio-MEG Decoding Network (AADN), which minimizes and maximizes mutual information to separate cross-modal consistency features from modality-specific features, thereby constructing a cross-modal consistency representation space. Additionally, by hierarchically integrating coarse-grained and fine-grained features, the model improves decoding accuracy and generalization capability. Experimental results show that, in a zero-shot audio-MEG cross-modal retrieval task involving 1,000 words, the framework achieves a top-1 accuracy of 43.1%, providing new insights into the neural mechanisms underlying speech perception.

Recently, contrastive learning has emerged as a promising approach for multi-view clustering (MVC), as it enforces cross-view consistency and leverages complementary information from different views to enhance the analysis of heterogeneous data. However, traditional contrastive MVC methods suffer from an inherent limitation: their one-to-many contrast mechanism induces the False Negative Problem (FNP), where semantically similar intra-class instances are erroneously repelled. This phenomenon compromises intra-class consistency and ultimately degrades clustering performance. To overcome this issue, we propose a novel Pseudo lA bel gU ided univerS um lE arning (PAUSE) framework for robust multi-view clustering. Specifically, PAUSE operates in two synergistic stages: (1) A warm-up stage that employs dual contrastive learning to generate reliable pseudo-labels, establishing robust semantic relationships; (2) A fine-tuning stage that synthesizes universum samples via Mixup between anchor instances and out-of-class centroids, guided by the acquired pseudo-labels. This unique mechanism constructs generalized negative classes that expand inter-class margins while preserving intra-class cohesion. Crucially, the widened decision boundaries prevent misclassification of displaced intra-class instances, effectively circumventing FNP without requiring explicit negative pair correction. We further devise a robust universum contrastive loss that explicitly enforces cross-view consistency through adaptive boundary constraints. Extensive experiments on five multi-view benchmarks demonstrate that our PAUSE consistently outperforms 11 state-of-the-art multi-view learning methods. Our code is accessible at: https://github.com/xixi-555/PAUSE_main_code.

Graph Attention Network (GAT) has a wide range of applications in HSI classification. The GAT-based semi-supervised learning approach enables the integration of valuable information from both labeled and unlabeled samples, effectively reducing the model’s reliance on labeled data. However, the node-wise training approach of GAT often overlooks the inherent global feature of graph data and the long-range dependencies among nodes, thereby limiting the model’s generalization ability on unlabeled data. Therefore, we propose a semi-supervised HSI classification model based on the dynamic evolution graph attention network (DEGAT). The main contributions: 1)We design a dynamic graph evolution mechanism (DGEM) that enables the model to capture the interactive information between local graph attention coefficients and the global graph structure, thus obtaining more discriminative graph representations. 2)DEGAT utilizes the multi-scale mechanism and message-passing mechanism to capture the information of nodes with long-range dependencies, extracting richer spatial-spectral features. State-of-the-art results are achieved with very few labeled training samples on two typical benchmark HSI datasets, where the overall accuracy reaches 95.12% and 98.76% respectively.

Convolutional neural networks (CNNs) have attained remarkable performance in hyperspectral image (HSI) classification. However, the existing CNNs are restricted by their limited receptive field in HSI classification. Recently, transformer networks have proved to be promising in many tasks thanks to the global receptive field, but they easily ignore some local information that is important for HSI classification. In this letter, we propose a novel method entitled convolutional transformer network (CTN) for HSI classification. In order to make full use of spectral information and spatial information, the method adopts center position encoding (CPE) to merge spectral features and pixel positions. Furthermore, the proposed method introduces convolutional transformer (CT) blocks. It effectively combines convolution and transformer structures together to capture local–global features of HSI patches, which is contributive for HSI classification. Experimental results on public datasets demonstrate the superiority of our method compared with several state-of-the-art classification methods. The codes of this work will be available athttps://github.com/sky8791to facilitate reproducibility.

Hyperspectral image (HSI) classification with a small number of training samples has been an urgently demanded task because collecting labeled samples for hyperspectral data is expensive and time-consuming. Recently, graph attention network (GAT) has shown promising performance by means of semisupervised learning. It combines the information of labeled and unlabeled samples so that the weakness of inadequate labeled samples is alleviated. In this letter, we propose a novel method, spectral–spatial GAT (SSGAT), for semisupervised HSI classification. The proposed SSGAT takes all samples (training and testing samples) as nodes and establishes an edge set among them to form a graph structure. In particular, the edge set is constructed in an unsupervised manner based on a large neighborhood to make full use of spectral–spatial information. Furthermore, the proposed method computes attention coefficients between a node and its neighbor nodes and aggregates them to generate more discriminative features, thus improving the performance of HSI classification. Experimental results on public data sets demonstrate the superiority of our proposed method compared with several state-of-the-art methods.

Convolutional neural network has been widely used in hyperspectral image classification. Compared with the early machine learning method, it has made great progress. However, the convolution kernel used in hyperspectral image classification ignores the intrinsic relationship among spatial pixels when extracting spectral features, which will lead to poor contour and very small false prediction in the classification results. Besides, The hyperspectral data can only be labeled by experts, which requires a lot of labor and material resources. In order to improve the classification accuracy of hyperspectral images and reduce the dependence on labeled samples, this paper proposed a hyperspectral image classification algorithm based on graph neural network. Through the characteristics of inherent points and edges in the graph, the spatial information and spectral information of hyperspectral images are fused. The features of unlabeled samples are used to participate in the training to improve the effect of classification model.

Hyperspectral images (HSIs) contain a large number of mixed pixels due to low spatial resolution, which poses great challenges to the analyses and applications of HSIs. In recent years, convolutional neural networks (CNNs) have attained promising performance in HSI field. However, few CNN-based methods are proposed to solve the hyperspectral unmixing (HU) problem because of insufficient labeled samples. In this paper, we propose a novel unsupervised method, sparsity constrained convolutional autoencoder network (SC-CAE), for the HU problem. The data are preprocessed by principal component analysis (PCA) and then fed into the encoder network to obtain low dimensional representations. The decoder network is to reconstruct the original data from these low dimensional representations. Under the sparse constraint, the endmember matrix and the abundance matrix are obtained after many training epochs. The experiment results on synthetic dataset and real dataset show that our method has evident advantages compared with several state-of-the-art methods.

Hyperspectral data contains abundant spectral do main information, which is of great significance to classification of objects. However, due to the lack of labeled data, it is difficult to get an acceptable result by just using the small number of labeled data. We propose a semi-supervised classification model based on convolutional neural network and introduce the attention mechanism to balance the sample weight. After the convolution of the multi-layer network, more information is concentrated on the channels, so we use the Squeeze-and-Excitation block, which can adaptively recalibrates the channel-wise characteristic response by explicitly modelling the inter-channel dependencies. At the same time, we used focal loss to reduce the problem of poor training caused by uneven samples. We test our model on hyperspectral data at Tianshan area. From the result, we can find that our method can get a great result on the mineral classification task, which can be used for making geological map.

Hyperspectral data contains abundant information in spectral domain, which is very useful for mineral classification and geological body mapping. But, due to the lack of labeled data, it is difficult to get an acceptable result by just using the small number of labeled data. We adopt a semi-supervised method called CNN, which can effectively extract inner features of hyperspectral image to classify hyperspectral data. However, with constraint to the size of receptive field, it can hardly get higher level features. We propose dilated CNN for mineral classification of hyperspectral data. At the same size of kernels, dilated CNN has bigger receptive field. At the meanwhile, it can get higher accuracy of classification. We test our model on hyperspectral data at Tianshan area, where is rich in minerals. From the result, we can find that our method can get a great result on the mineral classification task, which can be used for making geological map.

Classifying Hyperspectral images with few training samples is a challenging problem. The generative adversarial networks (GAN) are promising techniques to address the problems. GAN constructs an adversarial game between a discriminator and a generator. The generator generates samples that are not distinguishable by the discriminator, and the discriminator determines whether or not a sample is composed of real data. In this paper, by introducing multilayer features fusion in GAN and a dynamic neighborhood voting mechanism, a novel algorithm for HSIs classification based on 1-D GAN was proposed. Extracting and fusing multiple layers features in discriminator, and using a little labeled samples, we fine-tuned a new sample 1-D CNN spectral classifier for HSIs. In order to improve the accuracy of the classification, we proposed a dynamic neighborhood voting mechanism to classify the HSIs with spatial features. The obtained results show that the proposed models provide competitive results compared to the state-of-the-art methods.

Hyperspectral remote sensing data contains near continuous spectral information of the object, which is very suitable for mineral classification and geologic body mapping. However, the collecting of a lot of labeled hyperspectral data is expensive, time-consuming and labor-intensive. We choose a semi-supervised method to classify hyperspectral data based on a generative adversarial nertwork (GAN), just use a small amount of labeled data, named HSGAN. The GAN is made up of a generator and a discriminator, and the generator generates data similar to the real data so that the discriminator cannot tell if it is real data or generated data. We designed a one-dimensional GAN to extract spectral features from hyperspectral data. Using this method, we test the Tianshan hyperspectral data and use the actual geological map as the ground-truth produced by us. We find that HSGAN still achieves better results than the traditional CNN and SVM.

How to classify hyperspectral images using few training samples is an important and challenging problem because the collection of the samples is difficult and expensive. Because semi-supervised approaches can utilize information contained in the unlabeled samples and labeled samples, it is a suitable choice. A novel semi-supervised spectral-spatial classification method for hyperspectral data based on generative adversarial network (GAN) is proposed in this paper. First, we use a custom one-dimensional GAN to train the hyperspectral data to obtain spectral features. After using a new small convolutional neural network (CNN) to classify the spectral features, we use a new classification method based on a majority voting strategy further to improve the classification result. The performance of our method is evaluated on ROSIS image data, and the results show that the proposed method can acquire satisfactory results when compared with traditional methods using a few of labeled samples.

Because the collection of ground-truth labels is difficult, expensive, and time-consuming, classifying hyperspectral images (HSIs) with few training samples is a challenging problem. In this letter, we propose a novel semisupervised algorithm for the classification of hyperspectral data by training a customized generative adversarial network (GAN) for hyperspectral data. The GAN constructs an adversarial game between a discriminator and a generator. The generator generates samples that are not distinguishable by the discriminator, and the discriminator determines whether or not a sample is composed of real data. We design a semisupervised framework for HSI data based on a 1-D GAN (HSGAN). This framework enables the automatic extraction of spectral features for HSI classification. When HSGAN is trained using unlabeled hyperspectral data, the generator can generate hyperspectral samples that are similar to the real data, while the discriminator contains the features, which can be used to classify hyperspectral data with only a small number of labeled samples. The performance of the HSGAN is evaluated on the Airborne Visible Infrared Imaging Spectrometer image data, and the results show that the proposed framework achieves very promising results with a small number of labeled samples.

Band selection is a very important hyperspectral image preprocessing before using data. A novel bands selection method for hyperspectral data based on convolutional neural network (CNN) is proposed in this paper. In this way, we use a custom one-dimensional CNN to train the hyperspectral data to obtain a well-trained model. After testing band combinations, we use the model to obtain the test precision of the different band combinations, and finally use the band combination with the highest precision as the selected bands. This precision measure is a new criterion for band selection. This is the first application of CNN to band selection, and our proposed method can select the better combinations of band for specific problems. In the experiments, we select the bands on the Indian Pines dataset. The experimental results show that the proposed method can acquire satisfactory results when compared with traditional methods.

In this letter, a band-selection approach based on the deep convolutional neural network (CNN) and distance density (DD) is proposed. This method effectively mitigates the curse of dimensionality for hyperspectral images (HSIs). First, we use the hyperspectral full-band data to train a custom 1-D CNN to obtain a well-trained model. Second, we select band combinations based on DD. Using the rectified linear unit, which is the activation function of the CNN that is only activated with a nonzero value, we can effectively test the band combinations without retraining the model. Finally, the method selects the band combinations with the highest precision as the final selected bands. This precision measure is a new criterion for band selection. To further improve the performance, a data augmentation method based on DD is also proposed. To justify the effectiveness of the proposed method, experiments are conducted on two HSIs. The results show that the proposed method can acquire more satisfactory results than traditional methods.

Dependence degree (DD) and importance degree (ID) of patterns are crucial for pattern appraisement and model reconstruction. In rough set data analysis (RSDA), DD and ID lack robustness because the lower approximation set is terribly unstable under the indiscernibility relation perturbation. Statistical inference is a good way to deal with this instability. However, the fixed-value hypothesis testing and interval estimation of DD and ID were only discussed by the $\chi ^2$ test, which merely suits for two contingency tables (CT) with the same number of elements, and their nonzero elements must exist at the same positions. These requirements are too strict for practical applications. In this paper, the Kolmogorov-Smirnov (K-S) goodness-of-fit test is introduced to generalize the statistical inferences of DD and ID. As the bridge between data and corresponding measures, CT lies at the core of RSDA. By transforming CT to a random sample of a hypothesized random variable, the elementary algorithm for the goodness-of-fit test of contingency tables and the advanced algorithm for the goodness-of-fit test of contingency tables are proposed based on K-S statistic to implement goodness-of-fit tests of CTs. Better than $\chi ^2$ test, all CTs, even with different numbers and different positions of nonzero elements, are permitted. Subsequently, by generating a CT with expected DD value, the fixed-value hypothesis of DD is transformed to the goodness-of fit test between the original and expected CTs. Three algorithms, i.e., hypothesis test of dependence degree based on the K-S test, region estimation of the dependence degree, and significance test and region estimation of importance degree of attribute set, are proposed as fixed-value hypothesis tests and region estimations of DD and ID. These algorithms can be used to verify the importance of attributes and choose the attribute subset with the highest likelihood of maintaining the original discrimination ability. Experiments verify that the discrimination ability, disturbance tolerance ability, and stability under varying discretization strategies of DD and ID are significantly enhanced by the proposed algorithms.

The classic methods, such as FCM, often fail to carry out accurate modeling for the high-level fuzzy uncertainty, and then cause the classification error that should not be ignored in the application. Fortunately, the type-2 fuzzy set is a tool to handle this type of uncertainty. An adaptive interval-valued type-2 fuzzy C-Means clustering algorithm (A-IT2FCM) is proposed, including:(1) a proper modeling method for interval-valued type-2 fuzzy set;(2) an effective type reduction approach by adaptively searching the equivalent type-1 fuzzy sets for the type-2. Three different type-2 fuzzy clustering algorithms are used: the algorithm based on Karnik-Mendel type reduction, a method based on simple type reduction, and A-IT2FCM presented in this article. The experimental data are two data windows of SPOT5 image from Zhuhai and Beijing, China. Results show that, A-IT2FCM outperforms the other algorithms compared. Especially when obvious density difference exists between objects in the data, A-IT2FCM can achieve more accurate class boundaries and higher classification accuracy.

The pattern set of a remote sensing image contains many kinds of uncertainties. Uncertain information can create imperfect expressions for pattern sets in various pattern recognition algorithms, such as clustering algorithms. Methods based the fuzzy c-means algorithm can manage some uncertainties. As soft clustering methods, They are known to perform better on auto classification of remote sensing images than hard clustering methods. However, if the clusters in a pattern set are of different density and high order uncertainty, performance of FCM may significantly vary depending on the choice of fuzzifiers. Thus, we cannot obtain satisfactory results by using type-1 fuzzy set. Type-2 fuzzy sets permit us to model various uncertainties which cannot be appropriately managed by type-1 fuzzy sets. This paper introduces the theory of interval type-2 fuzzy set into the unsupervised classification of remote sensing images and proposes the automatic remote sensing image classification method based on the interval type-2 fuzzy c-means. Experimental results indicate that our method can obtain more coherent clusters and more accurate boundaries from the data with density difference. Our type-2 fuzzy model can manage the uncertainties of remote sensing images more appropriately and get a more desirable result.

This paper focus on two phenomena that "same spectrum with different objects" and "same object with different spectra" in multispectral remote sensing image, and propose a stepwise refinement classification method based on multi-sensitive strategies. It's a top-down, gradually refinement hierarchical way of classification which combines with advantages of both supervised classification and unsupervised classification: by analyzing the characteristic of spectrum curve, cluster and find out the band combinations with big characteristic differences as the guidance of classification; according to spectral characteristics of different bands combinations, choose different methods for further fine classification. Experimental results show that the proposed method achieved the high accurate classification of multispectral remote sensing images and effectively overcome the both above phenomena. Furthermore, the overall accuracy and kappa coefficient also confirm its superior performance.

To improve the quality of the fused image, we propose a remote sensing image fusion method based on sparse representation. In the method, first, we represent the source images with sparse coefficients. Second, the larger values of sparse coefficients of panchromatic (Pan) image is set to 0. Third, the coefficients of panchromatic (Pan) and multispectral (MS) image are combined with the linear weighted averaging fusion rule. Finally, the fused image is reconstructed from the combined sparse coefficients and the dictionary. The proposed method is compared with intensity-hue-saturation (IHS), Brovey transform (Brovey), discrete wavelet transform (DWT), principal component analysis (PCA) and fast discrete curvelet transform (FDCT) methods on several pairs of multifocus images. The experimental results demonstrate that the proposed approach performs better in both subjective and objective qualities.

The band selection of multispectral remote sensing image is a key issue and hot topic in remote sensing image processing domain. Considering that the band selection results of methods are usually not satisfying, a novel remote sensing image band selection method based on independent component analysis is proposed in this paper. The proposed method determine the number of independent components according to “Virtual dimension”, and then sort independent components, and select the top-ranking bands as important independent components. After that, sort the bands band by the contribution rate to important independent components, and finally we get the band combination results after remove the bands correlation. Statistical and visual results show that, the new proposed method can select the effectively and efficiently enhance the rock and soil information.

The remote sensing image classification is a key issue and hot topic in remote sensing image processing domain. Considering that the classification results of methods based on spectral angle or spectral distance are usually not satisfying, a novel remote sensing image classification method based on the combination of spectral angle and spectral distance is proposed in this paper. The proposed method utilizes the complementary of them to classify an image, that spectral angle is not sensitive to image gray. Moreover, based on the actual category of samples, weights of spectral angle and distance are automatically adjusted during the training process. Statistical and visual results show that, the proposed method is superior to methods respectively based on spectral angle and spectral distance in terms of visual effect, while overall classification accuracy and Kappa coefficient also confirm its superior performance.

Statistical inference about dependence degree (DD) and importance degree (ID) of variables in an information system is crucial for variables appraisement and model reconstruction. However, in rough set data analysis (RSDA), the literature is restricted to validate independence or test whether the degree is significantly big, while the fixed value test and interval estimation for related measurements have been ignored. Because these important issues have not been addressed, we cannot determine whether the data support expert opinions and compare the features in depth. To enhance the integrity of statistical inference for DD and ID in an RSDA, fixed value tests and interval estimations of DD and ID are presented in this paper. With multinomial distribution as the carrier for statistical information in the databases, the fixed value test of DD is successfully transformed into a restricted estimation of multinomial distribution and a goodness-of-fit test for distributions. The fixed value test and interval estimation algorithms for DD and ID are then presented in detail and illustrated with examples. Explicit expressions for the DD and ID interval estimation, DD confidence curves, and the limit theory for DD and ID are shown. Furthermore, the effectiveness and discrimination of the proposed algorithms are validated using the Car evaluation, Tic-Tac-Toe endgame, and Fisher's Iris databases.

This paper focus on eight frequently used Image fusion quality metrics (IFQMs), which are correlation coefficient, relative bias, structure similarity index, root-mean-square error, cross entropy, mutual information, spectral angle mapper and ERGAS to check their ability to measure the quality similarity among the images based on three different reference images. We have evaluated the performance of the IFQMs by considering in the three aspects: (1) Consistency with the other similar IFQMs; (2) Robustness to different testing images; (3) Consistency with the visual evaluations. Experimental results show that taking resampled multispectral image as reference, ERGAS outperforms other IFQMs. Taking original low resolution multi-spectral image as reference, correlation coefficient and SAM have the best performance. Taking the high resolution pan or SAR image as reference, root-mean-square error and ERGAS perform well. Relative bias is not suitable for fusion image evaluation due to its poor performance in all the three aspects.

Non-negative matrix factorization (NMF) is one of the recently emerged dimensionality reduction methods. Unlike other methods, NMF is based on non-negative constraints, which allows learn parts from objects. In this paper a performance comparison of PCA and NMF, which are data preprocessing algorithms in remote sensing imagery classification, is presented. PCA and NMF are applied to a remote sensing imagery (128× 128), obtained from Shunyi, Beijing. For classification, a maximum likelihood classification method is used for the preprocessed data. The results show that classification with NMF has more confident results than that with PCA. NMF keeps more abundant texture information.

In this paper, FastICA and KERNELICA algorithms and their application as preprocessing for remote sensing imagery (RSI) classification are discussed, as well as a comparison between the two algorithms. Both of the algorithms are applied to a TM RSI, obtained from Shunyi, Beijing, 1999. Then a Maximum Likelihood Classification (MLC) method uniting ISODATA is used to perform classification for the raw and preprocessed data. The results show that use of the data preprocessed gives more confident results than those obtained from the raw data. And for the two ICA algorithms, on one side, both are quite steady and can get rid of correlations existing in RSI. On the other side, KERNELICA appears to perform better than FastICA on texture information left in the independent component image (ICI) and classification accuracy. The convergence of the two

Linear estimation methods such as ordinary and simple kriging commonly fail to provide unbiased estimates of recovered ore tonnage and metal content which means that a mining project can be exposed to undue risk. Nonlinear estimation, such as the Gaussian disjunctive kriging (DK) technique provide a mean of calculating unbiased estimates of ore and metal content over any cut-off range and mining unit size combination. The disadvantage of this method is the requirement of an assumption of strict stationarity. It supposes that we have known all bivariate distributions of regional variables (Zalpha, Zbeta) and (Z0, Zbeta) in which the values of Zalpha, Zbetais known and the value of Z0is to be estimated. The paper contents of Gaussian anamorphosis, varigram and structure analysis and Hermite polynomials. The application of these estimation methods to a deposit is described. The study carries out DK and ordinary kriging (OK) for a suit of 3D drill data of a multi-metal deposit, containing 63 drills and 2 kinds of metal, from which we can see DK is more perfect and easier. A discussion of the results from a practical point of view is also given