Search

Heterogeneous palmprint recognition has attracted considerable research attention in recent years because it has the potential to greatly improve the recognition performance for personal authentication. In this article, we propose a simultaneous heterogeneous palmprint feature learning and encoding method for heterogeneous palmprint recognition. Unlike existing hand-crafted palmprint descriptors that usually extract features from raw pixels and require strong prior knowledge to design them, the proposed method automatically learns the discriminant binary codes from the informative direction convolution difference vectors of palmprint images. Differing from most heterogeneous palmprint descriptors that individually extract palmprint features from each modality, our method jointly learns the discriminant features from heterogeneous palmprint images so that the specific discriminant properties of different modalities can be better exploited. Furthermore, we present a general heterogeneous palmprint discriminative feature learning model to make the proposed method suitable for multiple heterogeneous palmprint recognition. Experimental results on the widely used PolyU multispectral palmprint database clearly demonstrate the effectiveness of the proposed method.

The superiority of Convolutional Neural Networks (CNNs) largely relies on their architectures that are usually manually crafted with extensive human expertise. Unfortunately, such kind of domain knowledge is not necessarily owned by every interested user. Data mining on existing CNNs can discover useful patterns and fundamental comments from their architectures, providing researchers with strong prior knowledge to design effective CNN architectures when they have no expertise in CNNs. There are various state-of-the-art data mining algorithms at hand, while there is only rare work on mining CNN architectures. One of the main reasons is the gap between CNN architectures and data mining algorithms. Specifically, the current CNN architecture descriptions cannot be exactly vectorized to feed to an data mining algorithm. In this paper, we propose a unified approach, named ArcText, to describing CNN architectures based on text. Particularly, four different units and an ordering method have been elaborately designed in ArcText, to uniquely describe the same CNN architecture with sufficient information. Also, the resulted description can be exactly converted back to the corresponding CNN architecture. ArcText bridges the gap between CNN architectures and data mining researchers, and has the potential to be utilized to wider scenarios.

Health wearables in combination with gamification enable interventions that have the potential to increase physical activity -- a key determinant of health. However, the extant literature does not provide conclusive evidence on the benefits of gamification, and there are persistent concerns that competition-based gamification approaches will only benefit those who are highly active at the expense of those who are sedentary. We investigate the effect of Fitbit leaderboards on the number of steps taken by the user. Using a unique data set of Fitbit wearable users, some of whom participate in a leaderboard, we find that leaderboards lead to a 370 (3.5%) step increase in the users' daily physical activity. However, we find that the benefits of leaderboards are highly heterogeneous. Surprisingly, we find that those who were highly active prior to adoption are hurt by leaderboards and walk 630 fewer steps daily after adoption (a 5% relative decrease). In contrast, those who were sedentary prior to adoption benefited substantially from leaderboards and walked an additional 1,300 steps daily after adoption (a 15% relative increase). We find that these effects emerge because sedentary individuals benefit even when leaderboards are small and when they do not rank first on them. In contrast, highly active individuals are harmed by smaller leaderboards and only see benefit when they rank highly on large leaderboards. We posit that this unexpected divergence in effects could be due to the underappreciated potential of noncompetition dynamics (e.g., changes in expectations for exercise) to benefit sedentary users, but harm more active ones., Management Science. Published online in Articles in Advance 19 Dec 2022. For supplemental material, please visit https://pubsonline.informs.org/doi/suppl/10.1287/mnsc.2022.4581

Spatially detailed characterization of the distribution amount and timing of river ice are important for identifying and predicting potential ice hazards. In this paper, we present an asynchronous river ice extraction and change detection method using multi-temporal SAR image and multi-spectral image. River channel information is a strong prior knowledge for ice detection and analysis. Therefore a river channel extraction algorithm on multi-spectral image based on sparse reconstruction is proposed and adopted in our method. The extracted river channel is used as prior information to effectively eliminate most interference regions on the shore. Then an adaptive threshold segmentation method is adopted to accurately detect river ice regions in SAR image. Fuzzy C-means clustering is used to segment river ice using the infrared bands of multi-spectral image, considering temperature can provide significant information to discriminate ice, water and shore. Finally, change analysis is done based on the ice extractions results of two kinds of images. The proposed method is applied on the Yellow River ice monitoring and experiments demonstrated that this straightforward approach works well with both SAR image and multi-spectral image.

Raindrops adhered to a glass window or camera lens appear in various blurring degrees and resolutions due to the difference in the degrees of raindrops aggregation. The removal of raindrops from a rainy image remains a challenging task because of the density and diversity of raindrops. The abundant location and blur level information are strong prior guide to the task of raindrop removal. However, existing methods use a binary mask to locate and estimate the raindrop with the value 1 (adhesion of raindrops) and 0 (no adhesion), which ignores the diversity of raindrops. Meanwhile, it is noticed that different scale versions of a rainy image have similar raindrop patterns, which makes it possible to employ such complementary information to represent raindrops. In this work, we first propose a soft mask with the value in [-1,1] indicating the blurring level of the raindrops on the background, and explore the positive effect of the blur degree attribute of raindrops on the task of raindrop removal. Secondly, we explore the multi-scale fusion representation for raindrops based on the deep features of the input multi-scale images. The framework is termed uncertainty guided multi-scale attention network (UMAN). Specifically, we construct a multi-scale pyramid structure and introduce an iterative mechanism to extract blur-level information about raindrops to guide the removal of raindrops at different scales. We further introduce the attention mechanism to fuse the input image with the blur-level information, which will highlight raindrop information and reduce the effects of redundant noise. Our proposed method is extensively evaluated on several benchmark datasets and obtains convincing results.

The small volume of the existing screen content images (SCIs) database with human ratings restricts the training processes of no-reference (NR) image quality assessment models based on traditional machine learning and deep learning. In this letter, we propose an NR model called the multi-task distortion-learning network to jointly analyse the distortion types and distortion degree of SCIs to be the prior knowledge for predicting the SCIs quality. Specifically, we first generate sufficient distorted SCIs labelled with the distortion type and degree, which does not need much effort to conduct subjective scoring experiments. Then, relying on these data, we pre-train a multi-task learning network to obtain strong prior knowledge about assessing the image quality. Finally, we further jointly train a quality assessment network with an attention module that simulates the mechanism of processing visual signals in the human eyes. The experimental results on the public SCIs databases show that the proposed model is competitive against other state-of-art approaches and achieves better consistency with the human vision system.

With the continuous spread of COVID-19 epidemic, the strict control of personnel makes it a problem to optimize the design of vehicle parameters after field measurement The energy absorption characteristics and deformation mode of the front structure of the vehicle determine the acceleration or force response of the vehicle body during the impact, which plays an important role in occupant protection The traditional multi-objective optimization method is to transform multi-objective problems into single objective optimization problems through weighted combination, objective planning, efficiency coefficient and other methods This method requires a strong prior knowledge The purpose of this paper is to combine the experimental design with the Multi-objective Particle Swarm Optimization (MPSO) method to achieve the optimization of the crash worthiness of automobile structure This method can effectively overcome the defect of low precision caused by the conventional response surface method in the whole design space In this paper, the multi-objective particle swarm optimization method is applied to the research of Crash worthiness optimization of automobile structure, which expands the application field of the multi-objective particle swarm optimization method, and also has a very big role in the optimization of other complex systems It can be seen from the experiment that the speed of multi-objective particle swarm optimization is much faster than that of other methods Only 100 iterations can get the relative better results The case study on the front structure of a car shows that the method has a good result It is of great significance to apply the method to the optimization design of the crash worthiness of the car structure to improve the crash safety of the car under the influence of COVID-19 epidemic © 2020 - IOS Press and the authors All rights reserved

Practical data sets often contain the label noise caused by various human factors or measurement errors, which means that a fraction of training examples might be mistakenly labeled. Such noisy labels will mislead the classifier training and severely decrease the classification performance. Existing approaches to handle this problem are usually developed through various surrogate loss functions under the framework of empirical risk minimization. However, they are only suitable for binary classification and also require strong prior knowledge. Therefore, this article treats the example features as side information and formulates the noisy label removal problem as a matrix recovery problem. We denote our proposed method as "label noise handling via side information" (LNSI). Specifically, the observed label matrix is decomposed as the sum of two parts, in which the first part reveals the true labels and can be obtained by conducting a low-rank mapping on the side information; and the second part captures the incorrect labels and is modeled by a row-sparse matrix. The merits of such formulation lie in three aspects: 1) the strong recovery ability of this strategy has been sufficiently demonstrated by intensive theoretical works on side information; 2) multi-class situations can be directly handled with the aid of learned projection matrix; and 3) only very weak assumptions are required for model design, making LNSI applicable to a wide range of practical problems. Moreover, we theoretically derive the generalization bound of LNSI and show that the expected classification error of LNSI is upper bounded. The experimental results on a variety of data sets including UCI benchmark data sets and practical data sets confirm the superiority of LNSI to state-of-the-art approaches on label noise handling.

Although hallucinations are not one of the Diagnostic and Statistical Manual of Mental Disorders-Fifth Edition (DSM-5) criteria for posttraumatic stress disorder (PTSD), they are increasingly documented in PTSD. They are noted in the absence of clear delusions, formal thought disorganization, disorganized speech, or behavior, ruling out a comorbid psychotic disorder like schizophrenia as a better explanation for these hallucinations. Hallucinations in both PTSD and schizophrenia share phenomenological features. We propose that hallucinations in PTSD, like those in schizophrenia, might be explained in terms of aberrant predictive coding, specifically the misapplication of strong prior beliefs that vitiate perceptual inference. This approach highlights the broader relationship between trauma and psychosis. Under predictive coding, the nervous system organizes past sensory data into an internal model of the world. Under stress, the brain prioritizes speed over accurate encoding. However, memories for traumatic experiences are typically strongly consolidated, to avoid similar experiences in future. In PTSD, this could lead to a world model comprised of inaccurate but overly precise prior beliefs, that can be triggered by stimuli tangentially related to the index trauma, resulting in hallucinations. Crucially, this evidence accumulation depends upon the relative precision of prior beliefs and sensory evidence (supplied in the form of prediction errors). Our basic argument is that stressful situations induce belief updating, in terms of precise prior beliefs, that are difficult to undo. These unduly precise, trauma-related beliefs then constitute perceptual hypotheses, memories, or narratives that bias subsequent experience. This prior bias may be so severe that sensory evidence is effectively ignored; that is, treated as very imprecise, in relation to prior beliefs. Such an account may lead to cognitive therapies for hallucinations aimed at strong prior beliefs, and the exciting prospect of combining such therapies with drugs that modulate neuroplasticity and enhance the adaptive consolidation of more appropriate priors. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Generating a realistic human class image from a sketch is a unique and challenging problem considering that the human body has a complex structure that must be preserved. Additionally, input sketches often lack important details that are crucial in the generation process, hence making the problem more complicated. In this article, we present an effective method for synthesizing realistic images from human sketches. Our framework incorporates human poses corresponding to locations of key semantic components (e.g., arm, eyes, nose), seeing that its a strong prior for generating human class images. Our sketch-image synthesis framework consists of three stages: semantic keypoint extraction, coarse image generation, and image refinement. First, we extract the semantic keypoints using Part Affinity Fields (PAFs) and a convolutional autoencoder. Then, we integrate the sketch with semantic keypoints to generate a coarse image of a human. Finally, in the image refinement stage, the coarse image is enhanced by a Generative Adversarial Network (GAN) that adopts an architecture carefully designed to avoid checkerboard artifacts and to generate photo-realistic results. We evaluate our method on 6,300 sketch-image pairs and show that our proposed method generates realistic images and compares favorably against state-of-the-art image synthesis methods.

In this paper, we propose an effective and efficient face deblurring algorithm by exploiting semantic cues via deep convolutional neural networks. As the human faces are highly structured and share unified facial components (e.g., eyes and mouths), such semantic information provides a strong prior for restoration. We incorporate face semantic labels as input priors and propose an adaptive structural loss to regularize facial local structures within an end-to-end deep convolutional neural network. Specifically, we first use a coarse deblurring network to reduce the motion blur on the input face image. We then adopt a parsing network to extract the semantic features from the coarse deblurred image. Finally, the fine deblurring network utilizes the semantic information to restore a clear face image. We train the network with perceptual and adversarial losses to generate photo-realistic results. The proposed method restores sharp images with more accurate facial features and details. Quantitative and qualitative evaluations demonstrate that the proposed face deblurring algorithm performs favorably against the state-of-the-art methods in terms of restoration quality, face recognition and execution speed.

This paper studies the information processing behavior of a decision maker (DM) who can only process a subset of all information he receives: before taking an action, the DM receives sequentially a number of signals and decides whether to process or ignore each of them as it is received. The model generates an information processing behavior consistent with that documented in the psychological literature: first, the DM chooses to process signals that are strong; second, his processing strategy exhibits confirmation bias if he has a strong prior belief; third, he tends to process signals that suggest favorable outcomes (wishful thinking). As an application I analyze how the Internet and the induced change in information availability affects the processing behavior of the DM. I show that providing more/better information to the DM could strengthen his confirming bias.

The complexity of orphan diseases, which are those that do not have an effective treatment, together with the high dimensionality of the genetic data used for their analysis and the high degree of uncertainty in the understanding of the mechanisms and genetic pathways which are involved in their development, motivate the use of advanced techniques of artificial intelligence and in-depth knowledge of molecular biology, which is crucial in order to find plausible solutions in drug design, including drug repositioning. Particularly, we show that the use of robust deep sampling methodologies of the altered genetics serves to obtain meaningful results and dramatically decreases the cost of research and development in drug design, influencing very positively the use of precision medicine and the outcomes in patients. The target-centric approach and the use of strong prior hypotheses that are not matched against reality (disease genetic data) are undoubtedly the cause of the high number of drug design failures and attrition rates. Sampling and prediction under uncertain conditions cannot be avoided in the development of precision medicine.

Hyper parameter optimization (HPO) is a crucial step in modern machine learning systems. Bayesian optimization (BO) has shown great promise in HPO, where the parameter evaluation is conducted through a black-box optimization procedure. However, the main drawback of BO lies in the expensive computation cost, which limits its application in large deep models. Hence there is strong motivation to reduce the number of epochs of training required by leveraging the partial information provided by iterative training procedures. Recent advancements use probabilistic models that extrapolate learning curves explicitly to early-stop the training of poor-performing configurations. However, these approaches impose a strong prior assumption on the distribution of learning curves and involve much larger computational complexity or rely heavily on predefined rules, which is not general for different cases. To tackle the challenge of training resource allocation in infinite parameter search space and in time horizon, we study HPO problem in Bayesian contextual bandits setting and derive several fully-dynamic strategies that are information-efficient and scalable. Extensive experiments demonstrate that the proposed method can significantly speed up the HPO process.

During postdiction, the last stimulus of a sequence changes the perception of the preceding ones: in the rabbit illusion, a three-flash series presented regularly in time, but not in space, is – illusory - perceived as spatially regular. Such a reorganization of (spatial) perception could be driven by internal priors, e.g. favoring slow motion for the rabbit illusion. Although postdiction is a ubiquitous phenomenon, its neural underpinnings remain poorly understood. Here, we focused on the role of priors during postdiction and hypothesized that these could be reflected by alpha oscillations (8-12Hz), previously observed to correlate with idiosyncratic biases. We presented human participants with ambiguous visual stimuli that elicited the rabbit illusion on about half the trials, allowing us to contrast MEG-EEG brain responses to the same physical events causing distinct percepts. Given that a strong prior will increase the overall probability of perceiving the illusion, we used the percentage of perceived illusion as a proxy for an individual’s prior. We found that high fronto-parietal alpha power was associated with perceiving the sequence according to individual biases: participants with high susceptibility to the illusion would report the illusion, while participants with low susceptibility would report the veridical sequence. Additionally, we found that pre-stimulus alpha phase in occipital and frontal areas dissociated illusory from non-illusory trials. These results point to a dissociated relation of the power and timing of alpha band activity to illusory perception, with power reflecting prior expectations and phase influencing behavioral performance, potentially due to the modulation of sensory uncertainty.Significance StatementLate events may sometimes influence how earlier events are perceived, as if the arrow of time was reversed in the brain. This surprising phenomenon, called postdiction, is observed in the rabbit illusion, and highlights a predominant mechanism for perceptual processes. Perception builds up from the combination of prior expectations with incoming sensory evidence, which takes time. We showed that pre-stimulus neural activity, and more specifically alpha oscillations (8-12Hz), play a double role in postdiction. Fronto-parietal alpha power reflects individual prior expectation, while occipital and frontal alpha phase predicts illusory perception. Postdiction might actually be a means of compensating for the neural delays inherent in perceptual processes, so that the arrow of perceptual time matches the arrow of physical time.

Images captured in snowy days suffer from noticeable degradation of scene visibility, which degenerates the performance of current vision-based intelligent systems. Removing snow from images thus is an important topic in computer vision. In this paper, we propose a Deep Dense Multi-Scale Network (DDMSNet) for snow removal by exploiting semantic and depth priors. As images captured in outdoor often share similar scenes and their visibility varies with depth from camera, such semantic and depth information provides a strong prior for snowy image restoration. We incorporate the semantic and depth maps as input and learn the semantic-aware and geometry-aware representation to remove snow. In particular, we first create a coarse network to remove snow from the input images. Then, the coarsely desnowed images are fed into another network to obtain the semantic and depth labels. Finally, we design a DDMSNet to learn semantic-aware and geometry-aware representation via a self-attention mechanism to produce the final clean images. Experiments evaluated on public synthetic and real-world snowy images verify the superiority of the proposed method, offering better results both quantitatively and qualitatively. https://github.com/HDCVLab/Deep-Dense-Multi-scale-Network https://github.com/HDCVLab/Deep-Dense-Multi-scale-Network .

Clustering single-cell RNA sequence (scRNA-seq) data poses statistical and computational challenges due to their high-dimensionality and data-sparsity, also known as `dropout' events. Recently, Regularized Auto-Encoder (RAE) based deep neural network models have achieved remarkable success in learning robust low-dimensional representations. The basic idea in RAEs is to learn a non-linear mapping from the high-dimensional data space to a low-dimensional latent space and vice-versa, simultaneously imposing a distributional prior on the latent space, which brings in a regularization effect. This paper argues that RAEs suffer from the infamous problem of bias-variance trade-off in their naive formulation. While a simple AE without a latent regularization results in data over-fitting, a very strong prior leads to under-representation and thus bad clustering. To address the above issues, we propose a modified RAE framework (called the scRAE) for effective clustering of the single-cell RNA sequencing data. scRAE consists of deterministic AE with a flexibly learnable prior generator network, which is jointly trained with the AE. This facilitates scRAE to trade-off better between the bias and variance in the latent space. We demonstrate the efficacy of the proposed method through extensive experimentation on several real-world single-cell Gene expression datasets., Comment: IEEE/ACM Transactions on Computational Biology and Bioinformatics

Heterogeneous graph embedding is to learn node representations that preserve both heterogeneity and semantic information of a heterogeneous graph, so as to facilitate various downstream tasks, such as node classification and link prediction. Most of the existing methods process the graph structure based on the meta-path to learn low-dimensional feature vectors. However, these methods need to manually set the meta-path which requires strong prior knowledge and omit the content features of the intermediate nodes along the meta-path. In this paper, we propose a novel heterogeneous graph neural network based on node position information (PHGNN). With this method, the meta-path is no longer required, and heterogeneous information can be automatically preserved. Specifically, the proposed framework consists of two main components: content-level learning and type-level learning. Content-level learning utilizes the features information of nodes to learn the importance of the edge-type based nodes to the target node, while type-level learning focuses on the importance of different types of nodes. With the two-level mechanism, PHGNN can hierarchically aggregate multi-type information. The experiments on multiple public datasets show that compared with state-of-the-art baselines, our proposed PHGNN model achieves significant and consistent improvements.

Automatic spine segmentation, based on ultrasound volume projection imaging (VPI), is of great value in clinical applications to diagnose scoliosis in teenagers. In this paper, we propose a novel framework to improve the segmentation accuracy on spine images via structure-enhanced attentive learning. Since the spine bones contain strong prior knowledge of their shapes and positions in ultrasound VPI images, we propose to encode this information into the semantic representations in an attentive manner. We first revisit the self-attention mechanism in representation learning, and then present a strategy to introduce the structural knowledge into the key representation in self-attention. By this means, the network explores both the contextual and structural information in the learned features, and consequently improves the segmentation accuracy. We conduct various experiments to demonstrate that our proposed method achieves promising performance on spine image segmentation, which shows great potential in clinical diagnosis.

Pedestrian tracking systems implemented in regular smartphones may provide a convenient mechanism for wayfinding and backtracking for people who are blind. However, virtually all existing studies only considered sighted participants, whose gait pattern may be different from that of blind walkers using a long cane or a dog guide. In this contribution, we present a comparative assessment of several algorithms using inertial sensors for pedestrian tracking, as applied to data from WeAllWalk, the only published inertial sensor dataset collected indoors from blind walkers. We consider two situations of interest. In the first situation, a map of the building is not available, in which case we assume that users walk in a network of corridors intersecting at 45° or 90°. We propose a new two-stage turn detector that, combined with an LSTM-based step counter, can robustly reconstruct the path traversed. We compare this with RoNIN, a state-of-the-art algorithm based on deep learning. In the second situation, a map is available, which provides a strong prior on the possible trajectories. For these situations, we experiment with particle filtering, with an additional clustering stage based on mean shift. Our results highlight the importance of training and testing inertial odometry systems for assisted navigation with data from blind walkers.

Coordinate-based neural representations have shown significant promise as an alternative to discrete, array-based representations for complex low dimensional signals. However, optimizing a coordinate-based network from randomly initialized weights for each new signal is inefficient. We propose applying standard meta-learning algorithms to learn the initial weight parameters for these fully-connected networks based on the underlying class of signals being represented (e.g., images of faces or 3D models of chairs). Despite requiring only a minor change in implementation, using these learned initial weights enables faster convergence during optimization and can serve as a strong prior over the signal class being modeled, resulting in better generalization when only partial observations of a given signal are available. We explore these benefits across a variety of tasks, including representing 2D images, reconstructing CT scans, and recovering 3D shapes and scenes from 2D image observations., Project page: https://www.matthewtancik.com/learnit

We present a versatile model, FaceAnime, for various video generation tasks from still images. Video generation from a single face image is an interesting problem and usually tackled by utilizing Generative Adversarial Networks (GANs) to integrate information from the input face image and a sequence of sparse facial landmarks. However, the generated face images usually suffer from quality loss, image distortion, identity change, and expression mismatching due to the weak representation capacity of the facial landmarks. In this paper, we propose to "imagine" a face video from a single face image according to the reconstructed 3D face dynamics, aiming to generate a realistic and identity-preserving face video, with precisely predicted pose and facial expression. The 3D dynamics reveal changes of the facial expression and motion, and can serve as a strong prior knowledge for guiding highly realistic face video generation. In particular, we explore face video prediction and exploit a well-designed 3D dynamic prediction network to predict a 3D dynamic sequence for a single face image. The 3D dynamics are then further rendered by the sparse texture mapping algorithm to recover structural details and sparse textures for generating face frames. Our model is versatile for various AR/VR and entertainment applications, such as face video retargeting and face video prediction. Superior experimental results have well demonstrated its effectiveness in generating high-fidelity, identity-preserving, and visually pleasant face video clips from a single source face image.

This paper offers a methodology to estimate an unconditional probability density function (PDF) for the stock price of an initial public offering (IPO), at a short-term post-IPO horizon. The resultant PDF is unique to the IPO of interest (IPOI) and serves to model the short-term post-market uncertainty associated with its price. Such a methodology is unprecedented in the IPO risk literature since the ex ante quantification of the short-term uncertainty associated with the stock price of a newly public firm was viewed as burdened by the lack of sufficient accounting and market history at the IPO stage. This gap is addressed here through recognizing that common in most IPO cases are the scarcity of hard data and abundance of soft data (strong prior belief), and that one can combine Bayesian inference and Data Envelopment Analysis (DEA) to develop a unique risk quantification setting that befits and serves these two characteristics of IPOs. In this setting, DEA serves to quantify the prior belief, to be subsequently updated in the Bayesian phase. This paper remains the first of its kind which unravels the IPO risk analysis from such perspective. It develops an iterative process that uses DEA to design a multi-dimensional similarity metric to find the ‘comparables’ to IPOI, and thereof the closest comparable to it, whereupon Bayesian inference is employed to utilize the information available from these comparables to sequentially update and revise the IPOI’s prior PDF. The validity of the proposed risk methodology was examined by backtesting analyses.

Traditional approaches to structural vector autoregressions (VARs) can be viewed as special cases of Bayesian inference arising from very strong prior beliefs. These methods can be generalized with a less restrictive formulation that incorporates uncertainty about the identifying assumptions themselves. We use this approach to revisit the importance of shocks to oil supply and demand. Supply disruptions turn out to be a bigger factor in historical oil price movements and inventory accumulation a smaller factor than implied by earlier estimates. Supply shocks lead to a reduction in global economic activity after a significant lag, whereas shocks to oil demand do not. (JEL C32, L71, Q35, Q43)

We use work histories retrospectively reported and matched to register data from the Swedish unemployment office to assess: 1) the prevalence of measurement error in reported spells of unemployment; 2) the impact of using such spells as the response variable of an exponential model; and 3) strategies for the adjustment of the measurement error. Due to the omission or misclassification of spells in work histories we cannot carry out typical adjustments for memory failures based on multiplicative models. Instead we suggest an adjustment method based on a mixture Bayesian model capable of differentiating between misdated spells and those for which the observed and true durations are unrelated. This adjustment is applied in two manners, one assuming access to a validation subsample and another relying on a strong prior for the mixture mechanism. Both solutions demonstrate a substantial reduction in the vast biases observed in the regression coefficients of the exponential model when survey data is used.

Background:Recently, mobile devices, such as smartphones, have been introduced into healthcare research to substitute paper diaries as data-collection tools in the home environment. Such devices support collecting patient data at different time points over a long period, resulting in clinical time-series data with high temporal complexity, such as time irregularities. Analysis of such time series poses new challenges for machine-learning techniques. The clinical context for the research discussed in this paper is home monitoring in chronic obstructive pulmonary disease (COPD).Objective:The goal of the present research is to find out which properties of temporal Bayesian network models allow to cope best with irregularly spaced multivariate clinical time-series data.Methods:Two mainstream temporal Bayesian network models of multivariate clinical time series are studied: dynamic Bayesian networks, where the system is described as a snapshot at discrete time points, and continuous time Bayesian networks, where transitions between states are modeled in continuous time. Their capability of learning from clinical time series that vary in nature are extensively studied. In order to compare the two temporal Bayesian network types for regularly and irregularly spaced time-series data, three typical ways of observing time-series data were investigated: (1) regularly spaced in time with a fixed rate; (2) irregularly spaced and missing completely at random at discrete time points; (3) irregularly spaced and missing at random at discrete time points. In addition, similar experiments were carried out using real-world COPD patient data where observations are unevenly spaced.Results:For regularly spaced time series, the dynamic Bayesian network models outperform the continuous time Bayesian networks. Similarly, if the data is missing completely at random, discrete-time models outperform continuous time models in most situations. For more realistic settings where data is not missing completely at random, the situation is more complicated. In simulation experiments, both models perform similarly if there is strong prior knowledge available about the missing data distribution. Otherwise, continuous time Bayesian networks perform better. In experiments with unevenly spaced real-world data, we surprisingly found that a dynamic Bayesian network where time is ignored performs similar to a continuous time Bayesian network.Conclusion:The results confirm conventional wisdom that discrete-time Bayesian networks are appropriate when learning from regularly spaced clinical time series. Similarly, we found that time series where the missingness occurs completely at random, dynamic Bayesian networks are an appropriate choice. However, for complex clinical time-series data that motivated this research, the continuous-time models are at least competitive and sometimes better than their discrete-time counterparts. Furthermore, continuous-time models provide additional benefits of being able to provide more fine-grained predictions than discrete-time models, which will be of practical relevance in clinical applications.

SUMMARYFull waveform inversion (FWI) is beginning to be used to characterize weak seismic events at different scales, an example of which is microseismic event (MSE) characterization. However, FWI with unknown sources is a severely underdetermined optimization problem, and hence requires strong prior information about the sources and/or the velocity model. The frequency-domain wavefield reconstruction inversion (WRI) method has shown promising results to mitigate the non-linearity of the FWI objective function generated by cycle-skipping. WRI relies on the reconstruction of data-assimilated wavefields, which approach the true wavefields near the receivers, a helpful feature when the source is added as an additional optimization variable. We present an adaptation of a recently proposed version of WRI based on the alternating direction method of multipliers that first finds the location of the MSEs and then reconstructs the wavefields and the source signatures jointly. Finally, the subsurface model is updated to focus the MSEs at their true locations. The method does not require prior knowledge of the number of MSEs. The inversion is stabilized by sparsifying regularizations separately tailored to the source location and velocity model subproblems. The method is tested on the Marmousi model using one MSE and two clusters of MSEs with two different initial velocity models, an accurate one and a rough one, as well as with added noise. In all cases, the method accurately locates the MSEs and recovers their source signatures.

3D reconstruction from a single image is a highly uncertain problem. Unknown information such as and depth of the self-occluded part of an object requires strong prior knowledge of the object. In order to make better use of prior information and generate high-quality 3D model, we first propose a 3D point cloud generator to obtain prior knowledge of the 3D point cloud in datasets. We then design a model to recover 3D shape of the object from a 2D image, which first estimates a 2.5D sketch from the input 2D image and then transfer the knowledge obtained in the 3D point cloud domain to the 2D image domain through the 2.5D sketch. Because 2.5D sketches are easier to obtain from 2D images and the information in a 2.5D sketch is more abundant than that in a 2D image Experiments demonstrate that our method is highly competitive to state-of-art works in 3D point cloud reconstruction on both synthetic datasets and real datasets.

As one of the common methods to construct classifiers, naïve Bayes has become one of the most popular classification methods because of its solid theoretical basis, strong prior knowledge learning characteristics, unique knowledge expression forms, and high classification accuracy. This classification method has a symmetry phenomenon in the process of data classification. Although the naïve Bayes classifier has high classification performance in single-label classification problems, it is worth studying whether the multilabel classification problem is still valid. In this paper, with the naïve Bayes classifier as the basic research object, in view of the naïve Bayes classification algorithm’s shortage of conditional independence assumptions and label class selection strategies, the characteristics of weighted naïve Bayes is given a better label classifier algorithm framework, the introduction of cultural algorithms to search for and determine the optimal weights is proposed as the weighted naïve Bayes multilabel classification algorithm. Experimental results show that the algorithm proposed in this paper is superior to other algorithms in classification performance.

For sequence transduction tasks like speech recognition, a strong structured prior model encodes rich information about the target space, implicitly ruling out invalid sequences by assigning them low probability. In this work, we propose local prior matching (LPM), a semi-supervised objective that distills knowledge from a strong prior (e.g. a language model) to provide learning signal to an end-to-end model trained on unlabeled speech. We demonstrate that LPM is simple to implement and superior to existing knowledge distillation techniques under comparable settings. Starting from a baseline trained on 100 hours of labeled speech, with an additional 360 hours of unlabeled data, LPM recovers 54%/82% and 73%/91% of the word error rate on clean and noisy test sets with/without language model rescoring relative to a fully supervised model on the same data.

Zero-shot learning (ZSL) aims to recognize novel classes by transferring semantic knowledge from seen classes to unseen classes. Since semantic knowledge is built on attributes shared between different classes, which are highly local, strong prior for localization of object attribute is beneficial for visual-semantic embedding. Interestingly, when recognizing unseen images, human would also automatically gaze at regions with certain semantic clue. Therefore, we introduce a novel goal-oriented gaze estimation module (GEM) to improve the discriminative attribute localization based on the class-level attributes for ZSL. We aim to predict the actual human gaze location to get the visual attention regions for recognizing a novel object guided by attribute description. Specifically, the task-dependent attention is learned with the goal-oriented GEM, and the global image features are simultaneously optimized with the regression of local attribute features. Experiments on three ZSL benchmarks, i.e., CUB, SUN and AWA2, show the superiority or competitiveness of our proposed method against the state-of-the-art ZSL methods. The ablation analysis on real gaze data CUB-VWSW also validates the benefits and accuracy of our gaze estimation module. This work implies the promising benefits of collecting human gaze dataset and automatic gaze estimation algorithms on high-level computer vision tasks. The code is available at https://github.com/osierboy/GEM-ZSL., Comment: Accepted by CVPR2021

In this paper, we propose 3D Context-Aware PIFu to recover 3D clothed human from a single image. Existing implicit function-based models suffer from the unsatisfied robustness to poor pose variations, since they ignore the inherent geometric relationship among 3D points. In this work, we utilize the 3D human model as a strong prior to regularize the reconstruction. With a fitted 3D human model, a global shape is extracted by the Pointnet to handle pose variations, and a local feature is extracted by Graph Convolutional Neural Network (GCNN) to capture geometry details. Besides, to enable the reconstruction network to capture fine-grained geometry on 3D cloth, we propose a multi-view implicit differentiable loss to directly measure the visual effect. Experimental results show that our approach is more robust to pose variations and reconstructs the human body with more details.

The visual examination of colonoscopic images fails to extract precise geometric information of the colonic surface. Reconstructing the 3D surface of the colon from colonoscopic image sequences may thus add valuable clinical information. We address this problem of extracting precise spatio-temporal 3D structure information from colonoscopic images. Using just the intrinsically calibrated monocular image stream, we develop a technique to compute the depth of certain feature points that have been tracked across images. Our method uses the prior knowledge of an approximate geometry of the colon, called the (TTP). It works by fitting a deformable cylindrical model to points reconstructed independently by non-rigid structure-from-motion (NRSfM), compromising between the data term and a novel tubular smoothing prior. Our method represents the first method ever to exploit a very weak topological prior to improve NRSfM. As such, it lies in-between standard NRSfM, which does not use a topological prior beyond the mere plane, and shape-from-template (SfT), which uses a very strong prior as a full deformable 3D object model. We validate our method on both synthetic images of tubular structures and real colonoscopic data. Our method improves the results obtained by existing NRSfM methods by 71.74% on average on synthetic data and succeeds in obtaining 3D reconstruction from a real colonoscopic sequence defeating the existing methods. Colonoscopic 3D reconstruction is a difficult problem, which is yet unresolved by the existing methods from computer vision. Our proposed dedicated NRSfM method and experiments show that the visual motion might be the right visual cue to use in colonoscopy.

This work investigates the problem of learning temporal interaction networks. A temporal interaction network consists of a series of chronological interactions between users and items. Previous methods tackle this problem by using different variants of recurrent neural networks to model interaction sequences, which fail to consider the structural information of temporal interaction networks and inevitably lead to sub-optimal results. To this end, we propose a novel Deep Structural Point Process termed as DSPP for learning temporal interaction networks. DSPP simultaneously incorporates the topological structure and long-range dependency structure into the intensity function to enhance model expressiveness. To be specific, by using the topological structure as a strong prior, we first design a topological fusion encoder to obtain node embeddings. An attentive shift encoder is then developed to learn the long-range dependency structure between users and items in continuous time. The proposed two modules enable our model to capture the user-item correlation and dynamic influence in temporal interaction networks. DSPP is evaluated on three real-world datasets for both tasks of item prediction and time prediction. Extensive experiments demonstrate that our model achieves consistent and significant improvements over state-of-the-art baselines.

Evaluation of music teaching is a highly subjective task often depending upon experts to assess both the technical and artistic characteristics of performance from the audio signal. This article explores the task of building computational models for evaluating music teaching using machine learning algorithms. As one of the widely used methods to build classifiers, the Naïve Bayes algorithm has become one of the most popular music teaching evaluation methods because of its strong prior knowledge, learning features, and high classification performance. In this article, we propose a music teaching evaluation model based on the weighted Naïve Bayes algorithm. Moreover, a weighted Bayesian classification incremental learning approach is employed to improve the efficiency of the music teaching evaluation system. Experimental results show that the algorithm proposed in this paper is superior to other algorithms in the context of music teaching evaluation.

Few-shot learning is to recognize novel concepts with few labeled samples. Recently, significant progress has been made to address the overfitting caused by data scarcity, especially those on modeling the distribution of novel categories given a single point. However, they often deeply rely on the prior knowledge from base set, which is generally hard to define, and its selection can easily bias the learning. A popular pipeline is to pretrain a feature extractor with base set and generate statistics from them as prior information. Since pretrained feature extractor cannot extract accurate representations for categories have never seen, and there is only 1 or 5 support images from novel categories, making it hard to acquire accurate priors, especially when they are far away from the class center. To address these issues, in this paper, we base our network on Maximum a posteriori (MAP), proposing a strategy for better prior selection from base set. We specially introduce semantic information, which are learned from unsupervised text corpora and easily available, to alleviate biases caused by unrepresentative support samples. Our intuition is that when the support from visual information is biased, semantics can provide strong prior knowledge to assist learning. Experimental results on four few-shot benchmarks also show that it outperforms the state-of-the-art methods by a large margin, improves around 2.08% ∼ 12.68% than the best results in each dataset on both 1- and 5-shot tasks.

Existing works address the problem of generating high frame-rate sharp videos by separately learning the frame deblurring and frame interpolation modules. Most of these approaches have a strong prior assumption that all the input frames are blurry whereas in a real-world setting, the quality of frames varies. Moreover, such approaches are trained to perform either of the two tasks - deblurring or interpolation - in isolation, while many practical situations call for both. Different from these works, we address a more realistic problem of high frame-rate sharp video synthesis with no prior assumption that input is always blurry. We introduce a novel architecture, Adaptive Latent Attention Network (ALANET), which synthesizes sharp high frame-rate videos with no prior knowledge of input frames being blurry or not, thereby performing the task of both deblurring and interpolation. We hypothesize that information from the latent representation of the consecutive frames can be utilized to generate optimized representations for both frame deblurring and frame interpolation. Specifically, we employ combination of self-attention and cross-attention module between consecutive frames in the latent space to generate optimized representation for each frame. The optimized representation learnt using these attention modules help the model to generate and interpolate sharp frames. Extensive experiments on standard datasets demonstrate that our method performs favorably against various state-of-the-art approaches, even though we tackle a much more difficult problem. The project page is available at https://agupt013.github.io/ALANET.html.

Existing techniques for dynamic scene reconstruction from multiple wide-baseline cameras primarily focus on reconstruction in controlled environments, with fixed calibrated cameras and strong prior constraints. This paper introduces a general approach to obtain a 4D representation of complex dynamic scenes from multi-view wide-baseline static or moving cameras without prior knowledge of the scene structure, appearance, or illumination. Contributions of the work are: An automatic method for initial coarse reconstruction to initialize joint estimation; Sparse-to-dense temporal correspondence integrated with joint multi-view segmentation and reconstruction to introduce temporal coherence; and a general robust approach for joint segmentation refinement and dense reconstruction of dynamic scenes by introducing shape constraint. Comparison with state-of-the-art approaches on a variety of complex indoor and outdoor scenes, demonstrates improved accuracy in both multi-view segmentation and dense reconstruction. This paper demonstrates unsupervised reconstruction of complete temporally coherent 4D scene models with improved non-rigid object segmentation and shape reconstruction and its application to free-viewpoint rendering and virtual reality., Submitted to IJCV 2019. arXiv admin note: substantial text overlap with arXiv:1603.03381

Determining the factors that shape temporal variation in species diversity is an ongoing challenge. One theory is that species exhibiting lower rates of phenotypic evolution should be more likely to go extinct as they are more susceptible to changing environmental conditions. However, little work has been done to assess whether this process shapes comparatively few lineages, or is a common mechanism shaping changes in species diversity. Here, I analyse the correlation between rates of morphological evolution and extinction at the species level using six published morphological matrices of non-avian dinosaurs. I find no correlation between the two rates at different taxonomic scales, suggesting that extinction in these groups is better described by other factors. As there is a strong prior expectation of correlated rates, I suggest that traditional morphological matrices are inappropriate for addressing this question and that the characters governing lineage persistence are independent of those with high phylogenetic signal. This may be comprehensively determined with continued development of phenomic matrices.

Humans expect downwards moving objects to accelerate and upwards moving objects to decelerate. These results have been interpreted as humans maintaining an internal model of gravity. We have previously suggested an interpretation of these results within a Bayesian framework of perception: earth gravity could be represented as a Strong Prior that overrules noisy sensory information (Likelihood) and therefore attracts the final percept (Posterior) very strongly. Based on this framework, we use published data from a timing task involving gravitational motion to determine the mean and the standard deviation of the Strong Earth Gravity Prior. To get its mean, we refine a model of mean timing errors we proposed in a previous paper (Björn Jörges & López-Moliner, 2019), while expands the range of conditions under which it yields adequate predictions of performance. This underscores our previous conclusion that the gravity prior is likely to be very close to 9.81 m/s². To obtain the standard deviation, we identify different sources of sensory and motor variability reflected in timing errors. We then model timing responses based on quantitative assumptions about these sensory and motor errors for a range of standard deviations of the earth gravity prior, and find that a standard deviation of around 2 m/s² makes for the best fit. This value is likely to represent an upper bound, as there are strong theoretical reasons along with support empirical evidence for the standard deviation of the earth gravity being lower than this value.

In this paper, we propose a learning method for deblurring Gaussian blurred images blindly by exploiting edge cues via deep multi-scales generative adversarial network: DeepEdgeGAN. We proposed the edges of the blurred images to be incorporated with the blurred image as the input of the DeepEdgeGAN to provide a strong prior constraint for the restoration, which is beneficial to solve the problem that gradients of the restored images with GANs methods tend to be smooth and not clear enough. Further, we introduce the perceptual and edge as well as scale losses to train the DeepEdgeGAN. With the trained end-to-end model, we directly restore the latent sharp images from blurred images and avoiding the estimation of pixel-kernel. Qualitative and quantitative experiments demonstrate that the visual effect of the restored images significantly improves better.

During thyroid ultrasound diagnosis, radiologists add markers such as pluses or crosses near a nodule's edge to indicate the location of a nodule. For computer-aided detection, deep learning models achieve classification, segmentation, and detection by learning the thyroid's texture in ultrasound images. Experiments show that manual markers are strong prior knowledge for data-driven deep learning models, which interferes with the judgment mechanism of computer-aided detection systems. Aiming at this problem, this paper proposes cascade marker removal algorithm for thyroid ultrasound images to eliminate the interference of manual markers. The algorithm consists of three parts. First, in order to highlight marked features, the algorithm extracts salient features in thyroid ultrasound images through feature extraction module. Secondly, mask correction module eliminates the interference of other features besides markers' features. Finally, the marker removal module removes markers without destroying the semantic information in thyroid ultrasound images. Experiments show that our algorithm enables classification, segmentation, and object detection models to focus on the learning of pathological tissue features. At the same time, compared with mainstream image inpainting algorithms, our algorithm shows better performance on thyroid ultrasound images. In summary, our algorithm is of great significance for improving the stability and performance of computer-aided detection systems. Graphical Abstract During thyroid ultrasound diagnosis, doctors add markers such as pluses or crosses near nodule's edge to indicate the location of nodule. Manual markers are strong prior knowledge for data-driven deep learning models, which interferes the judgment mechanism of computer-aided diagnosis system based on deep learning. Markers make models overfit the specific labeling forms easily, and performs poorly on unmarked thyroid ultrasound images. Aiming at this problem, this paper proposes a cascade marker removal algorithm to eliminate the interference of manual markers. Our algorithm make deep learning models pay attention on nodules' features of thyroid ultrasound images, which make computer-aided diagnosis system performs good in both marked imaging and unmarked imaging.

Siamese network based trackers have become a mainstream in visual object tracking. Recently, several high-performance multi-stage trackers have been proposed and some of them adopt SiamRPN for the first-stage region proposal. We argue that an anchor-based region proposal network is not necessary for the tracking task, as a tracker has a strong prior about the location and size of the target. In this paper, we propose a two-stage visual tracker which uses SiamFC for region proposal. SiamFC defines a bounding box by its center, which is a typical anchor-free (AF) network, so we dub our tracker AF2S. As the model size of SiamFC is only about 1/10 that of SiamRPN, AF2S results in a significantly lighter model than its SiamRPN-based counterparts. In the design of AF2S, we first build a strong AlexNet-based SiamFC baseline which improves the AUC on OTB-100 from 0.582 to 0.665. Further, we propose a position-sensitive convolutional layer which can be stacked after SiamFC backbone to increase the robustness of proposals without losing localization precision. Finally, a relation network is used for box refinement. Experimental results show that AF2S achieves the best performance on OTB-100 and VOT-18 among the state-of-the-art trackers which use AlexNet as backbone. On LaSOT-test, AF2S achieves an AUC of 0.480, which is among the first-tier performance even when trackers with more powerful backbone and much larger model size are considered.

Recent work has shown that the structure of convolutional neural networks (CNNs) induces a strong prior that favors natural images. This prior, known as a deep image prior (DIP), is an effective regularizer in inverse problems such as image denoising and inpainting. We extend the concept of the DIP to depth images. Given color images and noisy and incomplete target depth maps, we optimize a randomly-initialized CNN model to reconstruct a depth map restored by virtue of using the CNN network structure as a prior combined with a view-constrained photo-consistency loss. This loss is computed using images from a geometrically calibrated camera from nearby viewpoints. We apply this deep depth prior for inpainting and refining incomplete and noisy depth maps within both binocular and multi-view stereo pipelines. Our quantitative and qualitative evaluation shows that our refined depth maps are more accurate and complete, and after fusion, produces dense 3D models of higher quality.

We describe a new, original approach to the modelling of the Earth’s magnetic field. The overall objective of this study is to reliably render fast variations of the core field and its secular variation. This method combines a sequential modelling approach, a Kalman filter, and a correlation-based modelling step. Sources that most significantly contribute to the field measured at the surface of the Earth are modelled. Their separation is based on strong prior information on their spatial and temporal behaviours. We obtain a time series of model distributions which display behaviours similar to those of recent models based on more classic approaches, particularly at large temporal and spatial scales. Interesting new features and periodicities are visible in our models at smaller time and spatial scales. An important aspect of our method is to yield reliable error bars for all model parameters. These errors, however, are only as reliable as the description of the different sources and the prior information used are realistic. Finally, we used a slightly different version of our method to produce candidate models for the thirteenth edition of the International Geomagnetic Reference Field.

Generally, there are mainly two methods to solve the image restoration task in low-level computer vision, i.e., the model-based optimization method and the discriminative learning method. However, these two methods have clear advantages and disadvantages. For example, it is flexible for the model-based optimization method to handle different problems, but large quantity of computing time is required for better performance. The discriminative learning approach has high computing efficiency, but the application scope is seriously limited by the fixed training model. It would be better to combine the advantages of these two methods. Luckily, with the variable splitting techniques, we insert the trained convolutional neural network (CNN) for denoising as one model to the model-based optimization method to solve other image restoration problems (e.g., deblurring and super-resolution). Final experimental results show that our denoising network is able to provide strong prior information for image restoration tasks. The image restoration effects can reach or approximate the most advanced algorithm in such three tasks as denoising, deblurring, and super-resolution. Moreover, the algorithm proposed in this paper is also the most competitive in terms of computational efficiency.

CURRENT techniques for monocular reconstruction either require dense correspondences and rely on motion and deformation cues, or assume a highly accurate reconstruction (referred to as a template) of at least a single frame given in advance and operate in the manner of non-rigid tracking. Accurate computation of dense point tracks often requires multiple frames and can be computationally expensive. Availability of a template is a very strong prior which restricts system operation to a pre-defined environment and scenarios (see chapter 3.1 for the detailed introduction to the field and the state of the art).