Determination of petrophysical parameters is regarded as a critical task for the exploration and production of oil and gas reservoirs. Traditionally, the joint inversion of electromagnetic (EM) and seismic data under petrophysics constraints can be exploited to reconstruct the distribution of reservoir petrophysical parameters. However, methods involved with such schemes face challenges when solving high-contrast nonlinear inverse scattering problems due to the complexity of the relationship between resistivity/velocity and petrophysical properties and the nonuniqueness of these inverse problems. Here, to resolve these challenges, we have developed a field data transformation-joint inversion method (FDT-JIS) to directly reconstruct the distribution of porosity and water saturation. Specifically, the chain rule to transform geophysical parameters into petrophysical parameters is leveraged, and the field data transformation module is subsequently utilized to transform the scattered field data generated under the test configuration into those under the training configuration. Finally, a joint network is adopted to establish the mapping relationship between EM and acoustics data and petrophysical parameters to attain the inversion of petrophysical parameters. With numerical examples, we demonstrate that FDT-JIS not only allows different transceiver configurations to be used in training and testing, but also accurately reconstructs petrophysical parameters of complex models with high contrast in noisy environments.

In this work, a hybrid forward-inverse neural network (HFINN) with a transceiver transformation module (TTM) is proposed to increase the generalizability of machine learning-based electromagnetic (EM) inversion methods. The HFINN consists of two parts: a forward module and an inverse module. The inverse module combines the Res-Net with a fully convolutional network (FCN), which is called “Res-FCN,” to show the performance in wideband inversion; however, the data misfit from Res-FCN often remains high because it only minimizes the model misfit; thus, a forward module based on a classical neural network, U-Net, is trained first to alleviate the problem of large data misfit. To train HFINN better, a new loss function is proposed so that the frequency information is used as a prior physical constraint to optimize HFINN. Meanwhile, to further improve the generalizability of HFINN, TTM is incorporated into the HFINN as physical assistance so that it does not need to be retrained for different transceiver configurations. A total of 4000 random test samples are employed to verify the performance of the proposed HFINN, and the average model misfit is 27.15%. Six numerical examples are also provided to verify the inversion performance of HFINN over the whole frequency band. After adding the forward module, the average data misfit of the result is reduced by 7.5%. The numerical results show that HFINN performs well across the whole frequency band, even when the testing transceiver configuration is different from the training transceiver configuration.

In elastic wave inverse scattering problems, the material-property reconstruction, e.g. distributions of mass density, compressional wave speed, and shear wave speed from a limited set of measurement data, has attracted considerable research interest. However, simultaneous inversion of multiple parameters endures high computation costs, and reconstructed compressional and shear speeds may become unrealistic if no physical constraint is imposed. Meanwhile, because large objects with high contrasts over the background medium tend to induce high nonlinearity during the inversion process, it is difficult to obtain high-quality high-contrast material properties. To overcome such difficulties, we have developed a multi-module deep learning scheme with physical constraint for multi-parameter elastic wave inversion of high-contrast objects in inhomogeneous media. This scheme consists of (1) a preliminary imaging module (PIM), in which a deep residual network (ResNet) is employed to convert the scattered field data into the preliminary inversion images, (2) an image-enhancement module (IEM), in which a U-Net is employed to further enhance the image quality, and (3) a convolutional neural network (CNN) that is employed as the physical constraint module (PCM) to allow elastic wave parameters to satisfy actual physical constraint. Numerical examples have demonstrated that the proposed scheme not only accurately achieves multi-parameter elastic wave inversion, but also has good generalizability. Finally, the scheme can be applied to complex objects with high contrasts in both noise-free and noisy environments.

Recently, graph convolutional network (GCN) has received more and more interest in the field of hyperspectral image classification (HSIC). The existing GCN-based models for HSIC propagate and aggregate information through the GCN network based on the graph, which is constructed according to spatial location or spectral similarity. However, the constructed graph may not be ideal for the downstream classification task due to the variety of spectral characteristics. In this paper, a fully connected graph is adaptively constructed to make full use of local spatial information and global spectral information. Besides, we apply a neural sparsification technique to remove potentially task-irrelevant edges in case of misleading message propagation. Furthermore, label propagation (LP) serves as regularization to assist the graph network in learning proper edge weights that lead to improved classification performance. The resulting network is end-to-end trainable. The experimental results on three popular benchmarks, including Indian Pines, Pavia University, and Kennedy Space Center, demonstrate the superiority of our algorithm.

Graph neural network (GNN) has recently gained increasing attention in the hyperspectral image (HSI) classification. Compared with convolutional neural network (CNN), GNN can effectively relieve the scarcity of labeled data. In our method, we first perform feature learning on large-scale irregular regions through GNN and then extract local spatial–spectral features at the pixel level. Besides, we incorporate edge convolution (EdgeConv) into GNN to adaptively capture the interrelationship of the representative descriptors and fully exploit the discriminative features on graph. Experiments on several HSI datasets show that our method can achieve better classification performance compared with the state-of-the-art HSI classification methods.

Hyperspectral image (HSI) clustering is an extremely fundamental but challenging task with no labeled samples. Deep clustering methods have attracted increasing attention and have achieved remarkable success in HSI classification. However, most existing clustering methods are ineffective for large-scale HSI, due to their poor robustness, adaptability, and feature presentation. In this paper, to address these issues, we introduce unsupervised self-correlated learning smoothy enhanced locality preserving graph convolution embedding clustering (S2LGCC) for large-scale HSI. Specifically, the spectral-spatial transformation is introduced to transform the original HSI into a graph while preserving the local spectral features and spatial structures. After that, a locality preserving graph convolutional embedding encoder is designed to learn the hidden representation from the graph, in which the deep layer-wise graph convolutional network (LGAT) is proposed to preserve the adaptive layer-wise locality features. In addition, the self-correlated learning smoothy module is developed to learn the smoothy information and the non-local relationship in the hidden representation space for clustering. Finally, a self-training strategy is proposed to cluster the graph node, in which a self-training clustering objective employs soft labels to supervise the clustering process. The proposed S2LGCC is jointly optimized by the fusion graph reconstruction loss and self-training clustering loss, and the two benefit each other. On IP, Salinas, and UH2013 datasets, the OAs of our S2LGCC are 71.76%, 82.61%, and 63.82%, respectively.

Recently, with the explosive increase in Internet data, the traditional Graph-based Semi-Supervised Learning (GSSL) model is not suitable to deal with large scale data as the high computation complexity. Besides, GSSL models perform classification on a fixed input data graph. The quality of initialized graph has a great effect on the classification result. To solve this problem, in this paper, we propose a novel approach, named optimal bipartite graph-based SSL (OBGSSL). Instead of fixing the input data graph, we learn a new bipartite graph to make the result more robust. Based on the learned bipartite graph, the labels of the original data and anchors can be calculated simultaneously, which solves co-classification problem in SSL. Then, we use the label of anchor to handle out-of-sample problem, which preserves well classification performance and saves much time. The computational complexity of OBGSSL is O(ndmt+nm2), which is a significant improvement compared with traditional GSSL methods that need O(n2d+n3), where n, d, m and t are the number of samples, features anchors and iterations, respectively. Experimental results demonstrate the effectiveness and efficiency of our OBGSSL model.

In many real-world applications, we are often confronted with high dimensional data which are represented by various heterogeneous views. How to cluster this kind of data is still a challenging problem due to the curse of dimensionality and effectively integration of different views. To address this problem, we propose two parameter-free weighted multi-view projected clustering methods which perform structured graph learning and dimensionality reduction simultaneously. We can use the obtained structured graph directly to extract the clustering indicators, without performing other discretization procedures as previous graph-based clustering methods have to do. Moreover, two parameter-free strategies are adopted to learn an optimal weight for each view automatically, without introducing a regularization parameter as previous methods do. Extensive experiments on several public datasets demonstrate that the proposed methods outperform other state-of-the-art approaches and can be used more practically.

Recently, graph-based unsupervised feature selection algorithms (GUFS) have been shown to efficiently handle prevalent high-dimensional unlabeled data. One common drawback associated with existing graph-based approaches is that they tend to be time-consuming and in need of large storage, especially when faced with the increasing size of data. Research has started using anchors to accelerate graph-based learning model for feature selection, while the hard linear constraint between the data matrix and the lower-dimensional representation is usually overstrict in many applications. In this letter, we propose a flexible linearization model with anchor graph and [Formula: see text]-norm regularization, which can deal with large-scale data sets and improve the performance of the existing anchor-based method. In addition, the anchor-based graph Laplacian is constructed to characterize the manifold embedding structure by means of a parameter-free adaptive neighbor assignment strategy. An efficient iterative algorithm is developed to address the optimization problem, and we also prove the convergence of the algorithm. Experiments on several public data sets demonstrate the effectiveness and efficiency of the method we propose.