Prior Constraints Research Articles

Bayesian Opportunities for Brain–Computer Interfaces: Enhancement of the Existing Classification Algorithms and Out-of-Domain Detection

Bayesian neural networks (BNNs) are effective tools for a variety of tasks that allow for the estimation of the uncertainty of the model. As BNNs use prior constraints on parameters, they are better regularized and less prone to overfitting, which is a serious issue for brain–computer interfaces (BCIs), where typically only small training datasets are available. Here, we tested, on the BCI Competition IV 2a motor imagery dataset, if the performance of the widely used, effective neural network classifiers EEGNet and Shallow ConvNet can be improved by turning them into BNNs. Accuracy indeed was higher, at least for a BNN based on Shallow ConvNet with two of three tested prior distributions. We also assessed if BNN-based uncertainty estimation could be used as a tool for out-of-domain (OOD) data detection. The OOD detection worked well only in certain participants; however, we expect that further development of the method may make it work sufficiently well for practical applications.

Enhancing infrared imaging systems with temperature-dependent nonuniformity correction via single-frame and inter-frame structural similarity.

Temperature-dependent nonuniformity in infrared images significantly impacts image quality, necessitating effective solutions for intensity nonuniformity. Existing variational models primarily rely on gradient prior constraints from single-frame images, resulting in limitations due to insufficient exploitation of intensity characteristics in both single-frame and inter-frame images. This paper introduces what we believe to be a novel variational model for nonuniformity correction (NUC) that leverages single-frame and inter-frame structural similarity (SISB). This approach capitalizes on the structural similarities between the corrected image, intensity bias map, and degraded image, facilitating efficient suppression of intensity nonuniformity in real-world scenarios. The proposed method diverges fundamentally from existing strategies and demonstrates superior performance in comparison with state-of-the-art correction models.

Self supervised learning and the poverty of the stimulus

Diathesis alternations are the possible expressions of the arguments of verbs in different, systematically related subcategorization frames. Semantically similar verbs such as spill and spray can behave differently with respect to the alternations they can participate in. For example one can “spill/spray water on the plant”, but while one can “spray the plant with water”, it is odd to say “spill the plant with water”. “Spray” is a verb which can alternate between syntactic frames while “spill” is not alternating. How human speakers learn the difference between such verbs is not clearly understood, because the primary linguistic data (PLD) they receive does not appear sufficient to infer the knowledge required for adult competence. More generally the poverty of the stimulus (POS) hypothesis states that the PLD is not sufficient for a learner to infer full adult competence of language. That is, learning relies on prior constraints introduced by the language faculty. We tested state-of-the-art machine learning models trained by self supervision, and found some evidence that they could in fact learn the correct pattern of acceptability judgement in the locative alternation. However, we argued that this was partially a result of fine-tuning which introduced negative evidence into the learning data, which facilitated shortcut learning. Large language models (LLMs) cannot learn some linguistic facts from normal language data, but they can compensate to some extent by learning spurious correlated features when negative feedback is introduced during the training cycle.

Joint Direction of Arrival-Polarization Parameter Tracking Algorithm Based on Multi-Target Multi-Bernoulli Filter

This paper presents a tracking algorithm for joint estimation of direction of arrival (DOA) and polarization parameters, which exhibit dynamic behavior due to the movement of signal source carriers. The proposed algorithm addresses the challenge of real-time estimation in multi-target scenarios with an unknown number. This algorithm is built upon the Multi-target Multi-Bernoulli (MeMBer) filter algorithm, which makes use of a sensor array called Circular Orthogonal Double-Dipole (CODD). The algorithm begins by constructing a Minimum Description Length (MDL) principle, taking advantage of the characteristics of the polarization-sensitive array. This allows for adaptive estimation of the number of signal sources and facilitates the separation of the noise subspace. Subsequently, the joint parameter Multiple Signal Classification (MUSIC) spatial spectrum function is employed as the pseudo-likelihood function, overcoming the limitations imposed by unknown prior information constraints. To approximate the posterior distribution of MeMBer filters, Sequential Monte Carlo (SMC) method is utilized. The simulation results demonstrate that the proposed algorithm achieves excellent tracking accuracy in joint DOA-polarization parameter estimation, whether in scenarios with known or unknown numbers of signal sources. Moreover, the algorithm demonstrates robust tracking convergence even under low Signal-to-Noise Ratio (SNR) conditions.

D-VINS: Dynamic Adaptive Visual–Inertial SLAM with IMU Prior and Semantic Constraints in Dynamic Scenes

Visual–inertial SLAM algorithms empower robots to autonomously explore and navigate unknown scenes. However, most existing SLAM systems heavily rely on the assumption of static environments, making them ineffective when confronted with dynamic objects in the real world. To enhance the robustness and localization accuracy of SLAM systems in dynamic scenes, this paper introduces a visual–inertial SLAM framework that integrates semantic and geometric information, called D-VINS. This paper begins by presenting a method for dynamic object classification based on the current motion state of features, enabling the identification of temporary static features within the environment. Subsequently, a feature dynamic check module is devised, which utilizes inertial measurement unit (IMU) prior information and geometric constraints from adjacent frames to calculate dynamic factors. This module also validates the classification outcomes of the temporary static features. Finally, a dynamic adaptive bundle adjustment module is developed, utilizing the dynamic factors of the features to adjust their weights during the nonlinear optimization process. The proposed methodology is evaluated using both public datasets and a dataset created specifically for this study. The experimental results demonstrate that D-VINS stands as one of the most real-time, accurate, and robust systems for dynamic scenes, showcasing its effectiveness in challenging real-world scenes.

Azimuthal Prestack Seismic Anisotropic Inversion on a Deep and Tight Carbonate Reservoir From the North Potwar Basin of Pakistan

Summary Quantitative characterization of deep, tight, and heterogeneous reservoirs plays an important role in identifying hydrocarbon pathways for effective and optimal reservoir field development. In this case study, we used an azimuthal prestack seismic anisotropic inversion approach to estimate attributes of horizontal transverse isotropy (HTI) caused by a set of vertical fractures, oriented cracks, and stress. Anisotropic inversion facilitated the conversion of interface properties to the corresponding layer-based properties, which led to the quantitative interpretation of reservoir properties related to azimuthal variation in seismic amplitudes. To estimate the anisotropy magnitude and the direction of the isotropy axis in HTI media, elastic properties (P-impedance and Vp/Vs) obtained from prestack seismic inversion (using six azimuth × four angle stack) served as inputs. The isotropic low-frequency model (LFM) is used as the foundation of the inversion for all azimuths, and the anisotropy effects are later added by updating the model along the azimuths. The direction of the isotropy plane resulting from the anisotropic inversion is determined by using the maximum horizontal stress as a prior constraint, which eliminates any inherent uncertainty. The workflow used effectively characterized the orientation and density of fractures from the recently discovered oilfield reservoirs of the Paleocene (Lockhart) formation located in Pakistan’s north Potwar Basin. It also helped improve the prediction accuracy for fractures in the study area. According to the observations (fractures) from the exploratory drilled well (D1) in the tight carbonate (Lockhart) reservoir, a significant amount of anisotropy magnitude is observed. This provides the basis for hydrocarbon exploration, field development, and reliable drilling decisions.

Magnetic Resonance Image Denoising Based on Laplacian Prior Sparsity Constraint and Nonconvex Second-Order TV Penalty

Magnetic resonance (MR) imaging is considered as a very powerful imaging modality in clinical examination, but the process of image acquisition and transmission will be affected by noise, resulting in the degradation of imaging quality. In this paper, based on the Laplacian prior sparsity constraint and the nonconvex second order total variation (TV) penalty, we propose a MR images denoising model which consists of three terms. Specifically, in the first term, we use the L2-norm as the fidelity term to control the proximity between the observed image and the recovered MR image. Then, we introduce the Laplacian sparse prior constraint as the second term to mitigate the staircase artifacts in the recovered image. In the third term, we adopt the nonconvex second-order TV penalty to preserve important textures and edges. Finally, we use the alternating direction method of multipliers to solve the corresponding minimization problem. Comparative experiments on clinical data demonstrate the effectiveness of our approach in terms of PSNR and SSIM values.

An ionosphere-weighted calibration method with multi-station network corrections

Ionospheric delay is one of the most dominant error sources in GNSS fast relative positioning, especially for RTK positioning of medium and long baselines, which directly affects the ambiguity fixing successful rate and positioning accuracy. The strategy of combining ionospheric correction based on multi-reference station network with ionospheric weighted model is still one of the most effective methods for precise relative positioning. In this paper, a method of ionospheric weighted calibration model is presented. By using a large-scale reference station network at a certain period of time with ionospheric disturbance to obtain the prior ionospheric information required by the user, the GNSS precision relative positioning performance of the ionosphere-weighted calibration model in complex scenarios (the accuracy of prior ionospheric information will be reduced due to large-scale reference station network and ionospheric active scene) was systematically evaluated. The experimental results show that the ionosphere-weighted calibration (IWC) method can effectively reduce the influence of excessive ionospheric modeling residual on the positioning results, and significantly improve the fixed rate and positioning accuracy of RTK. Compared with the traditional ionosphere-fixed model based on virtual reference station (VRS) technology, the three-dimensional accuracy and fixed rate of the six rover stations are improved by 56% and 33.5% on average. Secondly, the prior ionospheric constraint intensity in IWC mode is controllable, which can effectively balance the relationship between initialization speed and positioning accuracy. On average, 94.3% of the time period of the six rover stations can complete the initialization and fixing within 10 epochs, which can basically meet the requirements of fast positioning. In addition, IWC model can ensure higher accuracy and better maintain fixed solution state after the ionospheric parameter convergence is completed. In the ionospheric disturbance stage, after a period of parameter convergence, high fixed rate and positioning accuracy can still be guaranteed.

Segmentation of Optic Disc and Optic Cup Based on Two-Layer Level Set with Sparse Shape Prior Constraint in Fundus Images

Glaucoma is a common high-incidence eye disease. The detection of the optic cup and optic disc in fundus images is one of the important steps in the clinical diagnosis of glaucoma. However, the fundus images are generally intensity inhomogeneity, and complex organizational structure, and are disturbed by blood vessels and lesions. In order to extract the optic disc and optic cup regions more accurately, we propose a segmentation method of the optic disc and optic cup in fundus image based on distance regularized two-layer level with sparse shape prior constraint. The experimental results show that our method can segment the optic disc and optic cup region more accurately and obtain satisfactory results.

B1 mapping using pre-learned subspaces for quantitative brain imaging.

To develop a machine learning-based method for estimation of both transmitter and receiver B1 fields desired for correction of the B1 inhomogeneity effects in quantitative brain imaging. A subspace model-based machine learning method was proposed for estimation of B1t and B1r fields. Probabilistic subspace models were used to capture scan-dependent variations in the B1 fields; the subspace basis and coefficient distributions were learned from pre-scanned training data. Estimation of the B1 fields for new experimental data was achieved by solving a linear optimization problem with prior distribution constraints. We evaluated the performance of the proposed method for B1 inhomogeneity correction in quantitative brain imaging scenarios, including T1 and proton density (PD) mapping from variable-flip-angle spoiled gradient-echo (SPGR) data as well as neurometabolic mapping from MRSI data, using phantom, healthy subject and brain tumor patient data. In both phantom and healthy subject data, the proposed method produced high-quality B1 maps. B1 correction on SPGR data using the estimated B1 maps produced significantly improved T1 and PD maps. In brain tumor patients, the proposed method produced more accurate and robust B1 estimation and correction results than conventional methods. The B1 maps were also applied to MRSI data from tumor patients and produced improved neurometabolite maps, with better separation between pathological and normal tissues. This work presents a novel method to estimate B1 variations using probabilistic subspace models and machine learning. The proposed method may make correction of B1 inhomogeneity effects more robust in practical applications.

Direct phasing algorithm for protein crystals with high solvent content using low-resolution diffraction data.

Over the past decade, iterative projection algorithms, an effective approach to recovering phases from a single intensity measurement, have found application in protein crystallography to directly surmount the `phase problem'. However, previous studies have always assumed that some prior knowledge constraints (i.e. a low-resolution envelope about the protein structure in the crystal cell or histogram matching requiring a similar density distribution to the target crystal) must be known for successful phase retrieval, thus hindering its widespread application. In this study, a novel phase-retrieval workflow is proposed that eliminates the need for a reference density distribution by utilizing low-resolution diffraction data in phasing algorithms. The approach involves randomly assigning one out of 12 possible phases at 30° intervals (or two for centric reflections) to produce an initial envelope, which is then refined through density modification after each run of phase retrieval. To evaluate the success of the phase-retrieval procedure, information entropy is introduced as a new metric. This approach was validated using ten protein structures with high solvent content, demonstrating its effectiveness and robustness.

Experimental determinations of carbon and hydrogen isotope fractionations and methane clumped isotope compositions associated with ethane pyrolysis from 550 to 600 °C

Methane clumped isotope compositions signify the relative natural abundances of rare, doubly substituted isotopic species of methane (13CH3D and 12CH2D2) and have emerged as a new isotopic tool to trace the sources, sinks, and lifecycles of methane in the environment. Such measurements can identify equilibration (or re-equilibration) temperatures if found to be in isotopic equilibrium or non-equilibrium processes (e.g., kinetically controlled reactions or mixing) if not in isotopic equilibrium. Naturally occurring thermogenic methane—formed by the thermally activated breakdown of larger organic molecules—has been found to have clumped isotope compositions consistent with equilibrium at reasonable gas formation temperatures in some settings and non-equilibrium processes occurring during either formation, migration, storage, or extraction in others. To explore the potential controls on the isotopic composition of thermogenic methane, we conducted isothermal time-series ethane pyrolysis experiments at 550 and 600 °C to measure methane and ethane 13C/12C and D/H fractionations and methane clumped isotope compositions (resolved 13CH3D and 12CH2D2). We explore the effects of modifying the initial clumped isotope composition of ethane and the addition of water vapor to pyrolysis experiments. We observe that ethane and methane 13C/12C are controlled by kinetic isotope effects and Rayleigh distillation processes. In contrast, ethane and methane D/H and methane clumped isotope compositions appear to be controlled by a combination of these processes and hydrogen isotope exchange. The hydrogen isotope exchange processes lead to isotopic equilibrium as reaction completion is approached for both D/H (ethane/methane) and methane clumped isotope compositions. We develop a chemical model based on a mass balance approach that accounts for inheritance vs. hydrogen-abstraction formation pathways for singly and doubly substituted isotopologues of ethane and methane that is compared to the experimental data. The model allows the determination of carbon and hydrogen kinetic isotope effects associated with ethane cracking and hydrogen abstraction reactions that, where applicable, we compare to prior theoretical constraints. From the comparison of the model to the experimental data, we infer that the kinetically controlled ethane and methane bulk isotope compositions and methane clumped isotope compositions are controlled by kinetic isotope effects (both primary and secondary) associated with both CC bond and CH bond cleavage reactions. Specifically, the methane clumped isotope compositions likely result from a combination of clumped isotope effects associated with ethane breakdown and/or assembly of methane isotopologues (expressed in terms of γ-factor parameters ≠ 1) and combinatorial effects that arise probabilistically. We discuss our experimental results in the context of recent pyrolysis experiments and observations of naturally occurring thermogenic methane. We consider a proposal consistent with observations from nature that the hydrogen isotope exchange reactions that promote equilibration of methane isotopic molecules at or near formation temperature may be facilitated by free radicals generated by pyrolysis reactions. In this framework, isotope exchange effectively ceases when pyrolysis effectively ceases locking in compositions that can be consistent with peak formation temperatures.

A Combined Deep Learning and Prior Knowledge Constraint Approach for Large-Scale Forest Disturbance Detection Using Time Series Remote Sensing Data

The scale and severity of forest disturbances across the globe are increasing due to climate change and human activities. Remote sensing analysis using time series data is a powerful approach for detecting large-scale forest disturbances and describing detailed forest dynamics. Various large-scale forest disturbance detection algorithms have been proposed, but most of them are only suitable for detecting high-magnitude forest disturbances (e.g., fire, harvest). Conversely, more continuous, subtle, and gradual lower-magnitude forest disturbances (e.g., thinning, pests, and diseases) have been subject to less focus. Deep learning (DL) can distinguish subtle differences in information within time series data, offering new opportunities to capture forest disturbances in a complete and detailed way. This study proposes an approach for analyzing forest dynamics across large areas and long time periods by combining DL time series classification and prior knowledge constraint. The approach consists of two stages: (1) an improved self-attention model used for time series classification to identify sequences with forest disturbance characteristics; (2) developed skip-disturbance recovery index (S-DRI) characterizing the temporal context, using prior knowledge constraint to identify forest disturbance years in time series with disturbance characteristics. In this study, the year of forest disturbances in five study areas located in the United States, Canada, and Poland from 2001 to 2020 was mapped. A total of 3082 manually interpreted test data with different disturbance causal agents (such as fire, harvest, conversion, hurricane, and pests) were sampled from five research areas for validation. Our approach was also evaluated against two forest disturbance benchmark datasets derived from LandTrendr and the Global Forest Change (GFC) dataset. The results demonstrate that our approach achieved an overall accuracy of 87.8%, surpassing the accuracy of LandTrendr (84.6%) and the Global Forest Change dataset (81.4%). Furthermore, our approach demonstrated lower omission rates (ranging from 10.0% to 67.4%) in detecting subtle to severe causal agents of forest disturbance, in comparison to LandTrendr (with a range of 18.0% to 81.6%) and GFC (with a range of 15.0% to 88.8%). This study, which involved mapping large-scale and long-term forest disturbance in multiple regions, revealed that our approach can be applied to new areas without a requirement for complex parameter adjustments. These results demonstrate the potential of our approach in generating comprehensive and detailed forest disturbance data, thus providing a new and effective method in this domain.

Urban Mobility and Knowledge Extraction from Chaotic Time Series Data: A Comparative Analysis for Uncovering COVID-19 Effects

The COVID-19 pandemic exerted devastating effects on the global economy, public health, and urban life. The geographical consequences lie in spatial time series dimensions, but the chaotic characteristics hinder intuitive understanding. Aiming to discover the variance of urban mobility in the pandemic context, this article presents a hybrid clustering technique called whole time sequence mixture (WTSM). The combination of whole time and subsequence techniques ensures robustness to data volume, dimensionality, sampling, distortion tasks, and a prior constraint. For yellow taxi trips in New York City, the case study spanned the pre-, mid-, and postpandemic periods and compared the performance of the established methods (symbolic aggregate approximation and dynamic time warping) and WTSM techniques. Findings revealed COVID-19 trends and social restriction-induced mobility variations and determined that vaccine supply did not lead to immediate mobility restoration. Meanwhile, concepts of validation indexes and computational complexity corroborate the superiority of WTSM. The transient cluster arising from temporal dissimilarity is a unique finding of WTSM, which led to high cluster cohesion and separation. WTSM can obtain new knowledge and rationale for urgent government intervention and epidemic management, not sacrificing computational efficiency and clustering quality. With the improved storage capability and desire for multidimensional pattern extraction, the increased accessibility to the chaotic time series data sets could promote further studies and help administrative schemes.

Sensing prior constraints in deep neural networks for solving exploration geophysical problems

One of the key objectives in geophysics is to characterize the subsurface through the process of analyzing and interpreting geophysical field data that are typically acquired at the surface. Data-driven deep learning methods have enormous potential for accelerating and simplifying the process but also face many challenges, including poor generalizability, weak interpretability, and physical inconsistency. We present three strategies for imposing domain knowledge constraints on deep neural networks (DNNs) to help address these challenges. The first strategy is to integrate constraints into data by generating synthetic training datasets through geological and geophysical forward modeling and properly encoding prior knowledge as part of the input fed into the DNNs. The second strategy is to design nontrainable custom layers of physical operators and preconditioners in the DNN architecture to modify or shape feature maps calculated within the network to make them consistent with the prior knowledge. The final strategy is to implement prior geological information and geophysical laws as regularization terms in loss functions for training the DNNs. We discuss the implementation of these strategies in detail and demonstrate their effectiveness by applying them to geophysical data processing, imaging, interpretation, and subsurface model building.

Combining Low-Light Scene Enhancement for Fast and Accurate Lane Detection

Lane detection is a crucial task in the field of autonomous driving, as it enables vehicles to safely navigate on the road by interpreting the high-level semantics of traffic signs. Unfortunately, lane detection is a challenging problem due to factors such as low-light conditions, occlusions, and lane line blurring. These factors increase the perplexity and indeterminacy of the lane features, making them hard to distinguish and segment. To tackle these challenges, we propose a method called low-light enhancement fast lane detection (LLFLD) that integrates the automatic low-light scene enhancement network (ALLE) with the lane detection network to improve lane detection performance under low-light conditions. Specifically, we first utilize the ALLE network to enhance the input image’s brightness and contrast while reducing excessive noise and color distortion. Then, we introduce symmetric feature flipping module (SFFM) and channel fusion self-attention mechanism (CFSAT) to the model, which refine the low-level features and utilize more abundant global contextual information, respectively. Moreover, we devise a novel structural loss function that leverages the inherent prior geometric constraints of lanes to optimize the detection results. We evaluate our method on the CULane dataset, a public benchmark for lane detection in various lighting conditions. Our experiments show that our approach surpasses other state of the arts in both daytime and nighttime settings, especially in low-light scenarios.

Green management, access to credit, and firms' vulnerability to the COVID-19 crisis.

The online version contains supplementary material available at 10.1007/s11187-023-00759-1.

AFLI-Calib: Robust LiDAR-IMU extrinsic self-calibration based on adaptive frame length LiDAR odometry

As an effective complement to common laser scanning systems, the portable laser scanning system can acquire point clouds flexibly and quickly. Calibration between Light detection and ranging (LiDAR) sensors and inertial measurement units (IMU) is the prerequisite for laser scanning systems to obtain high-quality point clouds. Related methods have been proposed in the last two decades, where the global navigation satellite system (GNSS) or high-precision calibration fields are commonly used. However, the extrinsic self-calibration of LiDAR-IMU is challenging, due to the large distortion in single-frame point cloud caused by rapid motion and the position errors of IMU integration which drift quickly. At the same time, the highly dynamic motion patterns of portable devices and the changes in the scanned scene structure are not well considered in existing LiDAR odometry methods. To take better advantage of the characteristics of non-repetitive scanning LiDAR sensor, this paper proposes AFLI-Calib, which utilizes adaptive frame length LiDAR odometry to perform the extrinsic self-calibration of LiDAR-IMU. Unlike LiDAR odometry methods with a fixed frame length, the LiDAR frame length is dynamically adjusted according to the motion state of sensors and the matching stability of scenes. The single-frame point cloud is registered to the map through a linear-based continuous-time model, eliminating the motion distortion correction in advance. For further optimization of trajectory and extrinsic parameters, IMU raw measurements and LiDAR observations are involved in the multi-constraint optimization, through tightly-coupled IMU pre-integration constraints, LiDAR point-to-plane constraints, and prior constraints. The method is fully validated using self-collected calibration data of indoor and outdoor scenes and different motion modes. Experiments show that on the test data, the translation parameter accuracy of the method is 0.041 m, which is 56.3% higher than the state-of-the-art method. The standard deviation is significantly reduced, with translation deviation (0.017 m, 0.024 m, 0.022 m) and rotation deviation (0.17°, 0.25°, 0.15°), which verifies the robustness of our method. The average RMSE of distances to the reference point cloud acquired by the terrestrial laser scanning system (TLS) is 0.042 m, showing a high accuracy calibration result. Comparative experiments with the fixed frame length LiDAR odometry method and classical “correction-then-registration” motion distortion model further verify the superiority and effectiveness of the proposed adaptive frame length LiDAR odometry.

The Elusive Progress of Prosodical Study: Essays in Honor of Thomas Cable

The Elusive Progress of Prosodical Study: Essays in Honor of Thomas Cable

Trade Conflict and Dynamic Changes in International Listed Company Stock Price: Evidence from Tesla

The occurrence of Sino-US trade war makes listed companies engaged in import and export trade bear two-way pull. On the one hand, the appreciation of the dollar to attract global capital into the U. S. markets, is a favorable contributing factor to rising stock prices; on the other hand, the trade protection policy of the United States has caused other countries to fight back. The increasing trade barriers have made the revenue of import and export enterprises decline, which is an unfavorable factor for stock prices. This paper selects Tesla, an international listed company, as the research object. The dynamic relationship between Tesla stock price and USD/RMB exchange rate is analyzed without any prior constraint, and then ARMA-GARCH model is constructed to simulate and predict stock return and volatility after introducing exchange rate as external explanatory variable, and based on the empirical results, puts forward reasonable and feasible suggestions for the development of international listed companies.