AbstractZellner's g prior remains a popular conventional for use in Bayesian variable selection, despite several undesirable consistency issues. In this article we study mixtures of priors as an alternative to default that resolve many the problems with original formulation while maintaining computational tractability has made so popular. We present theoretical properties mixture and provide real simulated examples compare fixed priors, empirical Bayes approaches, other procedures. Please...

This paper proposes a deep learning approach for accelerating magnetic resonance imaging (MRI) using large number of existing high quality MR images as the training datasets. An off-line convolutional neural network is designed and trained to identify mapping relationship between obtained from zero-filled fully-sampled k-space data. The not only capable restoring fine structures details but also compatible with online constrained reconstruction methods. Experimental results on real data have...

Abstract The hydrologic community has experienced a surge in interest machine learning recent years. This is primarily driven by rapidly growing data repositories, as well success of various academic and commercial applications, now possible due to increasing accessibility enabling hardware software. overview intended for readers new the field learning. It provides non‐technical introduction, placed within historical context, commonly used algorithms deep architectures. Applications sciences...

Linked or networked data are ubiquitous in many applications. Examples include web hypertext documents connected via hyperlinks, social networks user profiles friend links, co-authorship and citation information, blog data, movie reviews so on. In these datasets (called "information networks"), closely related objects that share the same properties interests form a community. For example, community blogsphere could be users mostly interested cell phone news. Outlier detection information can...

In this paper, we propose a new image representation to capture both the appearance and spatial information for classification applications. First, model feature vectors, from whole corpus, each at individual patch, in Bayesian hierarchical framework using mixtures of Gaussians. After such Gaussianization, is represented by Gaussian mixture (GMM) its appearance, several maps layout. Then extract GMM parameters, global local statistics over maps. Finally, employ supervised dimension reduction...

In recent years, diverse energy has been integrated into the power system, which constitutes a regional system (IES). However, coupling and complementation of multiple sources make load forecasting more difficult. For time-sequence non-linear characteristics electric complementarity different in IES, this paper proposed an attention-based convolutional neural network (CNN) combined with long short-term memory (LSTM) bidirectional (BiLSTM) model for IES. The historical load, temperature,...

Abstract Groundwater model structural error is ubiquitous, due to simplification and/or misrepresentation of real aquifer systems. During calibration, the basic hydrogeological parameters may be adjusted compensate for error. This result in biased predictions when such calibrated models are used forecast responses new forcing. We investigate impact on calibration and prediction a real‐world groundwater flow model, using Bayesian method with data‐driven explicitly account The error‐explicit...

Let $X| μ\sim N_p(μ,v_xI)$ and $Y| N_p(μ,v_yI)$ be independent p-dimensional multivariate normal vectors with common unknown mean $μ$. Based on only observing $X=x$, we consider the problem of obtaining a predictive density $\hat{p}(y| x)$ for $Y$ that is close to $p(y| μ)$ as measured by expected Kullback--Leibler loss. A natural procedure this (formal) Bayes $\hat{p}_{\mathrm{U}}(y| under uniform prior $π_{\mathrm{U}}(μ)\equiv 1$, which best invariant minimax. We show any will minimax if...

With the recent efforts made by computer vision researchers, more and types of features have been designed to describe various aspects visual characteristics. Modeling such heterogeneous has become an increasingly critical issue. In this paper, we propose a machinery called Heterogeneous Feature Machine (HFM) effectively solve recognition tasks in need multiple features. Our HFM builds kernel logistic regression model based on similarities that combine different distance metrics. Different...

We consider a Bayesian framework for estimating high-dimensional sparse precision matrix, in which adaptive shrinkage and sparsity are induced by mixture of Laplace priors. Besides discussing our formulation from the standpoint, we investigate MAP (maximum posteriori) estimator penalized likelihood perspective that gives rise to new nonconvex penalty approximating ℓ0 penalty. Optimal error rates estimation consistency terms various matrix norms along with selection structure recovery shown...

Kernel methods have been very popular in the machine learning literature last ten years, mainly context of Tikhonov regularization algorithms. In this paper we study a coherent Bayesian kernel model based on an integral operator defined as convolution with signed measure. Priors random measures correspond to prior distributions functions mapped by operator. We several classes and their image particular, identify general class whose is dense reproducing Hilbert space (RKHS) induced kernel. A...

The incorporation of unlabeled data in regression and classification analysis is an increasing focus the applied statistics machine learning literatures, with a number recent examples demonstrating potential for to contribute improved predictive accuracy. statistical basis this semisupervised does not appear have been well delineated; as result, underlying theory rationale may be underappreciated, especially by nonstatisticians. There also room statisticians become more fully engaged...

Abstract Effective water resources management typically relies on numerical models to analyze groundwater flow and solute transport processes. Groundwater are often subject input data uncertainty, as some inputs (such recharge well pumping rates) estimated uncertainty. Current practices of model calibration overlook uncertainties in data; this can lead biased parameter estimates compromised predictions. Through a synthetic case study surface‐ground interaction under changing conditions land...

We develop a supervised dimension reduction method that integrates the idea of localization from manifold learning with sliced inverse regression framework. call our localized (LSIR) since it takes into account local structure explanatory variables. The resulting projection LSIR is linear subspace variables captures nonlinear relevant to predicting response. applies both classification and problems can be easily extended incorporate ancillary unlabeled data in semi-supervised learning....

Fast and accurate depth estimation, or stereo matching, is essential in embedded vision systems, requiring substantial design effort to achieve an appropriate balance among accuracy, speed hardware cost. To reduce the right balance, we propose FP-Stereo for building high-performance matching pipelines on FPGAs automatically. consists of open-source hardware-efficient library, allowing designers obtain desired implementation instantly. Diverse methods are supported our library each stage...

This paper describes a novel method for edge feature detection of document images based on wavelet decomposition and reconstruction. By applying the technique, image becomes representation, i.e. is decomposed into set approximation coefficients detail coefficients. Discarding approximation, extraction implemented by means reconstruction technique. In consideration mutual frequency, overlapping will occur between details, multiresolution-edge with respect to an iterative procedure developed...

Kernel sliced inverse regression (KSIR) is a natural framework for nonlinear dimension reduction using the mapping induced by kernels. However, there are numeric, algorithmic, and conceptual subtleties in making method robust consistent. We apply two types of regularization this to address computational stability generalization performance. also provide an interpretation algorithm prove consistency. The utility approach illustrated on simulated real data.

In a remarkable series of papers beginning in 1956, Charles Stein set the stage for future development minimax shrinkage estimators multivariate normal mean under quadratic loss. More recently, parallel developments have seen emergence predictive densities Kullback–Leibler risk. We here describe these parallels emphasizing focus on Bayes procedures and derivation superharmonic conditions minimaxity as well further new density including multiple estimators, empirical linear model regression...

There has been an intense development on the estimation of a sparse regression coefficient vector in statistics, machine learning and related fields. In this paper, we focus Bayesian approach to problem, where sparsity is incorporated by so-called spike-and-slab prior coefficients. Instead replying MCMC for posterior inference, propose fast scalable algorithm based variational approximation distribution. The updating scheme employed our different from one proposed Carbonetto Stephens (2012)....

Invariance and low dimension of features are crucial significance in pattern recognition. This paper proposes a novel orthonormal shell Fourier descriptor that satisfies all these demands. method first performs decomposition on the line moment is obtained from 2-D pattern, then applies transform each scale coefficients. Unlike other existing wavelet-based methods, our allows applying common wavelets, such as Daubechies, Symmlet Coiflet, therefore it simple to implement. We study structure...