Graphical Modeling for Multi-Source Domain Adaptation Xu, Minghao; Wang, Hang; Ni, Bingbing
IEEE transactions on pattern analysis and machine intelligence,
2024-March, 2024-Mar, 2024-3-00, 20240301, Letnik:
46, Številka:
3
Journal Article
Recenzirano
Odprti dostop
Multi-Source Domain Adaptation (MSDA) focuses on transferring the knowledge from multiple source domains to the target domain, which is a more practical and challenging problem compared to the ...conventional single-source domain adaptation. In this problem, it is essential to model multiple source domains and target domain jointly, and an effective domain combination scheme is also highly required. The graphical structure among different domains is useful to tackle these challenges, in which the interdependency among various instances/categories can be effectively modeled. In this work, we propose two types of graphical models, i.e. C onditional R andom F ield for MSDA ( CRF-MSDA ) and M arkov R andom F ield for MSDA ( MRF-MSDA ), for cross-domain joint modeling and learnable domain combination. In a nutshell, given an observation set composed of a query sample and the semantic prototypes ( i.e. representative category embeddings) on various domains, the CRF-MSDA model seeks to learn the joint distribution of labels conditioned on the observations. We attain this goal by constructing a relational graph over all observations and conducting local message passing on it. By comparison, MRF-MSDA aims to model the joint distribution of observations over different Markov networks via an energy-based formulation, and it can naturally perform label prediction by summing the joint likelihoods over several specific networks. Compared to the CRF-MSDA counterpart, the MRF-MSDA model is more expressive and possesses lower computational cost. We evaluate these two models on four standard benchmark data sets of MSDA with distinct domain shift and data complexity, and both models achieve superior performance over existing methods on all benchmarks. In addition, the analytical studies illustrate the effect of different model components and provide insights about how the cross-domain joint modeling performs.
Gaussian likelihood inference has been studied and used extensively in both statistical theory and applications due to its simplicity. However, in practice, the assumption of Gaussianity is rarely ...met in the analysis of spatial data. In this paper, we study the effect of non‐Gaussianity on Gaussian likelihood inference for the parameters of the Matérn covariance model. By using Monte Carlo simulations, we generate spatial data from a Tukey g‐and‐h random field, a flexible trans‐Gaussian random field, with the Matérn covariance function, where g controls skewness and h controls tail heaviness. We use maximum likelihood based on the multivariate Gaussian distribution to estimate the parameters of the Matérn covariance function. We illustrate the effects of non‐Gaussianity of the data on the estimated covariance function by means of functional boxplots. Thanks to our tailored simulation design, a comparison of the maximum likelihood estimator under both the increasing and fixed domain asymptotics for spatial data is performed. We find that the maximum likelihood estimator based on Gaussian likelihood is overall satisfying and preferable than the non‐distribution‐based weighted least squares estimator for data from the Tukey g‐and‐h random field. We also present the result for Gaussian kriging based on Matérn covariance estimates with data from the Tukey g‐and‐h random field and observe an overall satisfactory performance.
The classical problem of peeling a beam off a substrate is studied through a re-examination of Griffith's fracture criterion in the presence of multiscale random properties. Four types of wide-sense ...homogeneous Gaussian random fields of the vector {Young's modulus E, surface energy density γ}, parametrized by the beam axis, are considered: Ornstein–Uhlenbeck, Matérn, Cauchy, and Dagum. The latter two are multiscale and allow decoupling of the fractal dimension and Hurst effects. Also calculated is the variance of the crack driving force G with any given type of random field in terms of the covariances of E and γ, under either fixed-grip or dead-load conditions. This investigation is complemented by a study of the stochastic crack stability which involves a stochastic competition between potential and surface energies. Overall, we find that, for Cauchy and Dagum models, the introduction of fractal-and-Hurst effects strongly influences the fracture mechanics results. Notably, while the Cauchy and Dagum models represent a more realistic scenario of random fields, given the same covariance on input, the response on output is strongest for Matérn, then Ornstein–Uhlenbeck, then Cauchy and, finally, Dagum model.
Semantic segmentation tasks can be well modeled by Markov Random Field (MRF). This paper addresses semantic segmentation by incorporating high-order relations and mixture of label contexts into MRF. ...Unlike previous works that optimized MRFs using iterative algorithm, we solve MRF by proposing a Convolutional Neural Network (CNN), namely Deep Parsing Network (DPN), which enables deterministic end-to-end computation in a single forward pass. Specifically, DPN extends a contemporary CNN to model unary terms and additional layers are devised to approximate the mean field (MF) algorithm for pairwise terms. It has several appealing properties. First, different from the recent works that required many iterations of MF during back-propagation, DPN is able to achieve high performance by approximating one iteration of MF. Second, DPN represents various types of pairwise terms, making many existing models as its special cases. Furthermore, pairwise terms in DPN provide a unified framework to encode rich contextual information in high-dimensional data, such as images and videos. Third, DPN makes MF easier to be parallelized and speeded up, thus enabling efficient inference. DPN is thoroughly evaluated on standard semantic image/video segmentation benchmarks, where a single DPN model yields state-of-the-art segmentation accuracies on PASCAL VOC 2012, Cityscapes dataset and CamVid dataset.
Synthetic Aperture Radar (SAR) image segmentation stands as a formidable research frontier within the domain of SAR image interpretation. The fully convolutional network (FCN) methods have recently ...brought remarkable improvements in SAR image segmentation. Nevertheless, these methods do not utilize the peculiarities of SAR images, leading to suboptimal segmentation accuracy. To address this issue, we rethink SAR image segmentation in terms of sequential information of transformers and cross-modal features. We first discuss the peculiarities of SAR images and extract the mean and texture features utilized as auxiliary features. The extraction of auxiliary features helps unearth the distinctive information in the SAR images. Afterward, a feature-enhanced FCN with the transformer encoder structure, termed FE-FCN, which can be extracted to context-level and pixel-level features. In FE-FCN, the features of a single-mode encoder are aligned and inserted into the model to explore the potential correspondence between modes. We also employ long skip connections to share each modality's distinguishing and particular features. Finally, we present the connection-enhanced conditional random field (CE-CRF) to capture the connection information of the image pixels. Since the CE-CRF utilizes the auxiliary features to enhance the reliability of the connection information, the segmentation results of FE-FCN are further optimized. Comparative experiments conducted on the Fangchenggang (FCG), Pucheng (PC), and Gaofen (GF) SAR datasets. Our method demonstrates superior segmentation accuracy compared to other conventional image segmentation methods, as confirmed by the experimental results.
•A simplified approach for generating conditional random field is proposed.•Analytical posterior statistics can be derived using the proposed approach.•The proposed approach is more efficient and ...accurate than adaptive Bayesian updating with structural reliability methods.•The actual spatial variation can be well characterized by conditional random field.•Borehole layout scheme affects the probability of slope failure significantly.
Conditional random field model can make best use of limited site investigation data to properly characterize the spatial variation of soil properties. This paper aims to propose a simplified approach for generating conditional random fields of soil undrained shear strength. A numerical method is adopted to validate the effectiveness of the proposed approach. With the proposed approach, the analytical posterior statistics of spatially varying undrained shear strength conditioned on the known values at measurement locations can be obtained. The conditional random field model of undrained shear strength is constructed using the field vane shear test data at a site of the west side highway in New York and the probability of slope failure is estimated by subset simulation. A clay slope under undrained conditions is investigated as an example to illustrate the proposed approach. The effects of borehole location and borehole layout scheme on the slope reliability are addressed. The results indicate that the proposed approach not only can well incorporate the limited site investigation data into modelling of the actual spatial variation of soil parameters by conditional random fields, but also can capture the depth-dependent nature of soil properties. The realizations of conditional random fields generated by the proposed approach can be well constrained to the site investigation data.
Random field theory is often used to model spatial variability of geo-material boundary and property. The results of random field generation based on different theories are quite different; however, ...few studies discuss the effects of adopting different random field approaches on the established stratigraphic models and their influence on engineering analysis. This article compares two random field approaches for evaluating liquefaction potential at a selected site. Here, based on the results of cone penetration tests (CPTs) at the study site, stratigraphic models are constructed using a continuous random field (conditional random field, CRF) and a discontinuous random field (Markov random field, MRF). Note that the MRF parameters were calibrated with the statistical parameters used in CRF. A series of geological profiles representing realizations of the derived CRF-based and MRF-based stratigraphic models are generated. Then, the liquefaction potential index (LPI) is calculated using the simplified procedure with a simulated geological profile and associated soil parameters. Finally, by repeating the analysis for all realizations of random stratigraphic models, the mean and the coefficient of variation of LPI are determined. Meanwhile, the uncertainty of stratigraphic models generated by CRF and MRF approaches is quantified and expressed as information entropy. Next, the relationship between stratigraphic model uncertainty (as an entropy) and LPI variation (or uncertainty) is established. The results show that: (1) the generation of the stratigraphic model is affected by the chosen random field approach, and the distribution of MRF-based strata is more continuous compared with that of CRF-based strata; (2) due to this effect, the strata uncertainty of CRF simulation is more uniform compared with that of MRF; (3) the information entropy and LPI uncertainty obtained using CRF exhibit moderate correlation, while these parameters obtained using MRF exhibit a strong positive correlation.
•This paper investigated the influence of geological (i.e., stratigraphic) uncertainty on the liquefaction potential index (LPI).•Stratigraphic models were created by two random field methods (conditional random field, CRF, and Markov random field, MRF).•The uncertainties of stratigraphic models were studied through the information entropy and average information entropy.•The variation of LPI was assessed by considering uncertainties of the stratigraphic model and the soil properties.•The relations of LPI variation and stratigraphic uncertainty from two approaches (CRF and MRF) are compared.
The random field finite element method (RF-FEM) provides a robust tool for carrying out slope reliability analysis that incorporates the spatial variability of soil properties. However, it has a ...major drawback of being computationally very time-consuming. To address this common criticism, the current study proposes a novel metamodel-based method for efficient slope reliability analysis in spatially variable soils. The proposed method involves the use of Convolutional Neural Networks (CNNs) as metamodels of the random field finite element model. With proper training using a small but sufficient number of random field samples, the CNN can potentially replace the computationally demanding random field finite element analyses for Monte-Carlo simulations. This paper examines the capability of CNNs to learn high-level features that contain information about the random variabilities in both spatial distribution and intensity, and the accuracy of the subsequent predictions of the RF-FEM results. Application of the proposed method to slope reliability analysis in spatially variable soils is illustrated and compared against other metamodel-based approaches, using a case study involving a multi-layered soil system with randomly varying cohesion c and the friction angle ϕ. The results show that (i) the proposed CNN approach predicts a probability of slope failure that is within 5% of the corresponding value obtained using direct RF-FEM Monte-Carlo simulations, but at a small fraction of the computational cost, and (ii) the proposed method also compares favourably against other metamodel-based methods in terms of computational efficiency and accuracy.
•Deep-learning is introduced for slope reliability analysis in spatially variable soils.•Random fields are configured in an image-like manner for processing using Convolutional Neural Networks (CNNs).•CNNs are capable of learning information pertaining to the random variabilities in both spatial distribution and intensity.•CNNs successfully provide accurate regressions between information about the random variabilities and the slope's factor of safety.•The proposed approach is validated and compared against other approaches.
Site characterization, which aims to characterize the subsurface stratigraphic configuration and the associated geo-properties, has long been a significant challenge in geological and geotechnical ...practice. Due to the complexity and inherent spatial variability of the geological bodies and the limited availability of borehole data, uncertainty is unavoidable in the characterized subsurface stratigraphic configuration and the associated geo-properties. In previous studies, the stratigraphic uncertainty and the geo-properties uncertainty were characterized separately. This paper proposes a conditional random field approach for a coupled characterization of stratigraphic and geo-properties uncertainties. The spatial correlation of the stratum existence between different subsurface elements and the spatial correlation of geo-properties are characterized by two autocorrelation functions, determined with the maximum likelihood principle. With the knowledge of the spatial correlation of the stratum existence, the stratigraphic configuration can be sampled using a modified random field approach. Then, the spatial correlation of the geo-properties is updated based on the sampled stratigraphic configuration. With the updated spatial correlation of the geo-properties, the spatial distribution of the geo-properties can readily be simulated with the conditional random field theory. The effectiveness of the proposed approach is demonstrated through a case study of probabilistic site characterization of an offshore wind farm site in Taiwan. To extend the applicability of the proposed approach, a probabilistic evaluation of liquefaction potential at this site under a given seismic shaking level is performed.
•A probabilistic method for characterizing the geological model is proposed.•Characterizations of stratigraphic and geo-properties uncertainties are coupled.•Stratigraphic configuration is simulated with modified random field approach.•Spatial variability of geo-properties is simulated with conditional random field.•The spatial correlation structure is estimated with maximum likelihood principle.