Prior Information Research Articles

Repulsion, chaos, and equilibrium in mixture models

Abstract Mixture models are commonly used in applications with heterogeneity and overdispersion in the population, as they allow the identification of subpopulations. In the Bayesian framework, this entails the specification of suitable prior distributions for the weights and locations of the mixture. Despite their popularity, the flexibility of these models often does not translate into the interpretability of the clusters. To overcome this issue, repulsive mixture models have been recently proposed. The basic idea is to include a repulsive term in the distribution of the atoms of the mixture, favouring mixture locations far apart. This approach induces well-separated clusters, aiding the interpretation of the results. However, these models are usually not easy to handle due to unknown normalizing constants. We exploit results from equilibrium statistical mechanics, where the molecular chaos hypothesis implies that nearby particles spread out over time. In particular, we exploit the connection between random matrix theory and statistical mechanics and propose a novel class of repulsive prior distributions based on Gibbs measures associated with joint distributions of eigenvalues of random matrices. The proposed framework greatly simplifies computations thanks to the availability of the normalizing constant in closed form. We investigate the theoretical properties and clustering performance of the proposed distributions.

CPDet: Circle-Permutation-Aware Object Detection for Heat Exchanger Cleaning

Shell–tube heat exchangers are commonly used equipment in large-scale industrial systems of wastewater heat exchange to reclaim the thermal energy generated during industrial processes. However, the internal surfaces of the heat exchanger tubes often accumulate fouling, which subsequently reduces their heat transfer efficiency. Therefore, regular cleaning is essential. We aim to detect circle holes on the end surface of the heat exchange tubes to further achieve automated positioning and cleaning tubes. Notably, these holes exhibit a regular distribution. To this end, we propose a circle-permutation-aware object detector for heat exchanger cleaning to sufficiently exploit prior information of the original inputs. Specifically, the interval prior to the extraction module extracts interval information among circle holes based on prior statistics, yielding prior interval context. The following interval prior fusion module slices original images into circle domain and background domain maps according to the prior interval context. For the circle domain map, prior-guided sparse attention using prior a circle–hole diameter as the step divides the circle domain map into patches and performs patch-wise self-attention. The background domain map is multiplied by a hyperparameter weak coefficient matrix. In this way, our method fully leverages prior information to selectively weigh the original inputs to achieve more effective hole detection. In addition, to adapt the hole shape, we adopt the circle representation instead of the rectangle one. Extensive experiments demonstrate that our method achieves state-of-the-art performance and significantly boosts the YOLOv8 baseline by 5.24% mAP50 and 5.25% mAP50:95.

Fan noise radial mode detection based on [formula omitted] regularization method

Radial mode detection has been widely employed in fan noise research. Many radial mode detection algorithms by conducting the convex L1 norm optimization have been proposed by taking advantage of the fact that the fan noise sound field is usually dominated by a limited set of radial modes. Nevertheless, L1 norm optimization has been shown to be suboptimal and cannot enforce further sparsity. A fan noise radial mode detection method based on L1/2 regularization algorithm is proposed in this paper to improve its robustness and dynamic range. This method employs the L1/2 regularization to directly solve the in-duct sound propagation model. Synthesized simulations indicates that the L1/2 regularization method can detect the correct radial mode order and amplitude under low signal-to-noise ratio. Moreover, the new method is insensitive to regularization parameter. An experimental system of radial rake is implemented to validate the new method. It is shown that the L-curve method, which do not require any prior information of the acoustic mode distribution or contaminating measurement noise, is an appropriate choice of the regularization parameter determination method for the L1/2 regularizer. It is demonstrated that the new L1/2 regularization method can enhance the mode sparsity, especially under low signal-to-noise ratio. Moreover, the resolution and dynamic range can be improved by the L1/2 regularization method, which makes it an effective and robust fan noise radial mode detection technique in practical applications.

Crossfeat: a transformer-based cross-feature learning model for predicting drug side effect frequency.

Safe drug treatment requires an understanding of the potential side effects. Identifying the frequency of drug side effects can reduce the risks associated with drug use. However, existing computational methods for predicting drug side effect frequencies heavily depend on known drug side effect frequency information. Consequently, these methods face challenges when predicting the side effect frequencies of new drugs. Although a few methods can predict the side effect frequencies of new drugs, they exhibit unreliable performance owing to the exclusion of drug-side effect relationships. This study proposed CrossFeat, a model based on convolutional neural network-transformer architecture with cross-feature learning that can predict the occurrence and frequency of drug side effects for new drugs, even in the absence of information regarding drug-side effect relationships. CrossFeat facilitates the concurrent learning of drugs and side effect information within its transformer architecture. This simultaneous exchange of information enables drugs to learn about their associated side effects, while side effects concurrently acquire information about the respective drugs. Such bidirectional learning allows for the comprehensive integration of drug and side effect knowledge. Our five-fold cross-validation experiments demonstrated that CrossFeat outperforms existing studies in predicting side effect frequencies for new drugs without prior knowledge. Our model offers a promising approach for predicting the drug side effect frequencies, particularly for new drugs where prior information is limited. CrossFeat's superior performance in cross-validation experiments, along with evidence from case studies and ablation experiments, highlights its effectiveness.

Sparse-Coding Variational Autoencoders.

The sparse coding model posits that the visual system has evolved to efficiently code natural stimuli using a sparse set of features from an overcomplete dictionary. The original sparse coding model suffered from two key limitations, however: (1) computing the neural response to an image patch required minimizing a nonlinear objective function via recurrent dynamics and (2) fitting relied on approximate inference methods that ignored uncertainty. Although subsequent work has developed several methods to overcome these obstacles, we propose a novel solution inspired by the variational autoencoder (VAE) framework. We introduce the sparse coding variational autoencoder (SVAE), which augments the sparse coding model with a probabilistic recognition model parameterized by a deep neural network. This recognition model provides a neurally plausible feedforward implementation for the mapping from image patches to neural activities and enables a principled method for fitting the sparse coding model to data via maximization of the evidence lower bound (ELBO). The SVAE differs from standard VAEs in three key respects: the latent representation is overcomplete (there are more latent dimensions than image pixels), the prior is sparse or heavy-tailed instead of gaussian, and the decoder network is a linear projection instead of a deep network. We fit the SVAE to natural image data under different assumed prior distributions and show that it obtains higher test performance than previous fitting methods. Finally, we examine the response properties of the recognition network and show that it captures important nonlinear properties of neurons in the early visual pathway.

A prior knowledge-embedded deformable convolutional network for vision-based fabric defect detection

The vision-based defect detection of textile surface is an essential problem in evaluating the appearance quality. In previous studies on vision-based defect detection, the following two points were neglected: (1) The proportion of defects on fabric is small, resulting in a low signal-to-noise ratio of defect images and a lack of sampling features; (2) the irregular shape of defects can overlap in position, leading to localization difficulty. In this paper, we propose a prior knowledge-embedded deformable convolutional network (PKE-DCNet) based on deep learning to address these two issues. First, a feature extraction method with prior information is designed with a shape-matching deep clustering module and a region-biased sampling module for detecting defects with complex shape. Then, a defect boundary boxes adaptive generation method is proposed with an anchor-free search mechanism and irregular edge contour computation to detect the key points in the region. Extensive experiments with mixed fabric defect datasets demonstrated that PKE-DCNet reached an overall mAP of 95.36% for six types of defects within a detection speed of 322 FPS, which was better than state-of-the-art methods.

Adaptive control of cardiac rhythms

Cardiac rhythms are related to heart electrical activity, being the essential aspect of the cardiovascular physiology. Usually, these rhythms are represented by electrocardiograms (ECGs) that are useful to detect cardiac pathologies. Essentially, the heart activity starts in the sinoatrial node (SA) node, the natural pacemaker, propagating to the atrioventricular node (AV), and finally reaching the His-Purkinje complex (HP). This paper investigates the control of cardiac rhythms in order to induce normal rhythms from pathological responses. A mathematical model that presents close agreement with experimental measurements is employed to represent the heart functioning. The adopted model comprises a network of three nonlinear oscillators that represent each one of the cardiac nodes, connected by delayed couplings. The pathological behavior is induced by an external stimulus in the SA node. An adaptive controller is proposed acting in the SA node considering an strategy based on the signal obtained by the natural pacemaker and its regularization. The incorporation of adaptive compensation in a Lyapunov-based control scheme allows the compensation for the unknown dynamics. The controller ability to deal with interpatient variability is evaluated by assuming that the heart model is not available to the controller design, being used only in the simulator to assess the control performance. Results show that the adaptive term can reduce the control effort by around 3% while reducing the tracking error by 20%, when compared to the conventional feedback approach. Additionally, the controller can avoid abnormal rhythms, turning the ECG closer to the expected normal behavior and preventing critical cardiac responses. Therefore, this work demonstrates that an adaptive controller can be used to regulate the ECG signal without prior information about the system and disregarding inter- and intrapatient variability.

A cautionary note against selective applications of the Bayes factor.

Bayes factor analysis becomes increasingly popular, among other reasons, because it allows to provide evidence for the null hypothesis, which is not easily possible with the traditional frequentist approach. A conceivable strategy that apparently takes favorable aspects of both approaches on board is to use traditional frequentist analyses first and to support theoretically interesting nil effects by Bayesian analyses thereafter. Here, we asked whether such a selective application of Bayesian analyses to only nonsignificant effects of foregoing frequentist analyses creates bias. In two simulation studies, we observed that such selective application of Bayesian analyses, in fact, severely overestimates evidence in favor of the null hypotheses, when a true population effect exists. While this bias can be attenuated by using more informative priors in the Bayesian analyses, we recommend to not apply such selective combination of analytical approaches, but instead to use either frequentist or Bayesian analyses consistently. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Pharmacometrics-Enhanced Bayesian Borrowing for Pediatric Extrapolation - A Case Study of the DINAMO Trial.

Bayesian borrowing analyses have an important role in the design and analysis of pediatric trials. This paper describes use of a prespecified Pharmacometrics Enhanced Bayesian Borrowing (PEBB) analysis that was conducted to overcome an expectation for reduced statistical power in the pediatric DINAMO trial due to a greater than expected variability in the primary endpoint. The DINAMO trial assessed the efficacy and safety of an empagliflozin dosing regimen versus placebo and linagliptin versus placebo on glycemic control (change in HbA1c over 26 weeks) in young people with type 2 diabetes (T2D). Previously fitted pharmacokinetic and exposure-response models for empagliflozin and linagliptin based on available historical data in adult and pediatric patients with T2D were used to simulate participant data and derive the informative component of a Bayesian robust mixture prior distribution. External experts and representatives from the U.S. Food and Drug Administration provided recommendations to determine the effective sample size of the prior and the weight of the informative prior component. Separate exposure response-based Bayesian borrowing analyses for empagliflozin and linagliptin showed posterior mean and 95% credible intervals that were consistent with the trial results. Sensitivity analyses with a full range of alternative weights were also performed. The use of PEBB in this analysis combined advantages of mechanistic modeling of pharmacometric differences between adults and young people with T2D, with advantages of partial extrapolation through Bayesian dynamic borrowing. Our findings suggest that the described PEBB approach is a promising option to optimize the power for future pediatric trials.

Using scenario tree modelling to evaluate the probability of freedom from Enzootic bovine leukosis (EBL) in Italy and Slovenia.

Documented freedom from disease is paramount for international free trade of animals and animal products. This study describes a scenario tree analysis to estimate the probability of freedom from Enzootic bovine leukosis (EBL) in Italy and Slovenia using information gathered via the data collection tool developed in the COST action project SOUND-control. Data on EBL control programmes (CPs) from 2018 to 2021 were used to build the models. Since animals are only sampled on the farm, one surveillance system component (SSC) was considered. The posterior probability of freedom (PostPfree) was estimated in time steps of one year, from 2018 to 2021. After each year, the calculated from the previous year, combined with the probability of introduction, was used as a prior probability for the next year. The herd level design prevalence was set to 0.2% in accordance with the Council Directive 64/432/EEC and the within herd design prevalence was set to 15%. As Slovenia implemented a risk-based surveillance, targeting the herds importing cattle, in its model the design herd prevalence was combined with an average adjusted risk to calculate the effective probability of a herd importing cattle being infected. The models were run for 10,000 iterations. Over the study period the mean estimates were: i) for Italy both the surveillance system sensitivity ( SSe) and PostPFree 100%, with no differences between simulations and years, ii) for Slovenia the SSe was 50.5% while the PostPFreewas 81.6%.

Laplacian-enhanced non-synchronous measurements method

To address the challenges of physical geometric constraints of microphone arrays and the lack of a prior information about the operating frequencies of target signals, we propose a novel method for broadband multiple-sound-source localization: a non-synchronous measurements method for broadband multiple-sound-source localization based on Laplacian-enhanced low-rank tensor completion. Our method thoroughly analyzes the tensor data structure of broadband signals and introduces a multiplier-optimized alternating direction method. Through extensive simulation studies, our approach has achieved remarkable results. We validate the outstanding performance of the Laplacian-enhanced algorithm, demonstrating its ability to generate highly accurate sound maps, capturing five distinct speech signal sources. Compared to other methods, our approach provides a more comprehensive global view. This technological advancement brings substantial hope for the precise localization of multiple broadband sound sources in critical applications, offering significant potential for future applications.

A graph-based method for automatic graded diagnosis of depression using EEG signals

Depression is a common psychological disorder typically categorized into four grades according to the symptom severity. Accurately diagnosing different grades of depression is significant for taking appropriate treatment strategies to prevent the condition from worsening. Electroencephalography (EEG) has the potential to mirror the prevailing mental state of patients, providing a possible way to diagnose depression. However, it is still a challenge to extract EEG features that reflect varying degrees of depression for automatic grading diagnosis. It has been pointed out that depression involves the intensification of abnormal interactions among distinct brain regions as it advances, which suggests that brain spatial information may play a key role in tracking different grades of depression. Based on this, combined with deep learning models, this paper proposes a new graph-based method for automatic grading diagnosis of depression. A novel graph-based brain network (BN) construction method is first proposed to convert EEG into graph structure containing brain spatial information, where the individual-level (IBN) and group-level brain networks (GBN) are constructed respectively. Next, a new deep learning model P-I-GAT is built based on graph attention networks (GAT) for identifying EEG signals from depression patients with different severity. In specific, the “P” represents the prior information integration capability, synthesizing attribute information from the IBN and structural information from the GBN. The resulting synthesized brain network contains prior information from these two different scales, referred to as the prior brain network (PBN). “I” denotes the multi-rhythm interaction capability, which adaptively assigns learning weights to each rhythmic PBN to fully exploit the information embedded within the PBNs. Ultimately, the proposed method is validated on the open dataset MODMA, achieving an accuracy of 88.57% and a specificity of 91.97%.

Radiation image reconstruction and uncertainty quantification using a Gaussian process prior

We propose a complete framework for Bayesian image reconstruction and uncertainty quantification based on a Gaussian process prior (GPP) to overcome limitations of maximum likelihood expectation maximization (ML-EM) image reconstruction algorithm. The prior distribution is constructed with a zero-mean Gaussian process (GP) with a choice of a covariance function, and a link function is used to map the Gaussian process to an image. Unlike many other maximum a posteriori approaches, our method offers highly interpretable hyperparamters that are selected automatically with the empirical Bayes method. Furthermore, the GP covariance function can be modified to incorporate a priori structural priors, enabling multi-modality imaging or contextual data fusion. Lastly, we illustrate that our approach lends itself to Bayesian uncertainty quantification techniques, such as the preconditioned Crank–Nicolson method and the Laplace approximation. The proposed framework is general and can be employed in most radiation image reconstruction problems, and we demonstrate it with simulated free-moving single detector radiation source imaging scenarios. We compare the reconstruction results from GPP and ML-EM, and show that the proposed method can significantly improve the image quality over ML-EM, all the while providing greater understanding of the source distribution via the uncertainty quantification capability. Furthermore, significant improvement of the image quality by incorporating a structural prior is illustrated.

Point Clouds Matching Based on Discrete Optimal Transport.

Matching is an important prerequisite for point clouds registration, which is to establish a reliable correspondence between two point clouds. This paper aims to improve recent theoretical and algorithmic results on discrete optimal transport (DOT), since it lacks robustness for the point clouds matching problems with large-scale affine or even nonlinear transformation. We first consider the importance of the used prior probability for accurate matching and give some theoretical analysis. Then, to solve the point clouds matching problems with complex deformation and noise, we propose an improved DOT model, which introduces an orthogonal matrix and a diagonal matrix into the classical DOT model. To enhance its capability of dealing with cases with outliers, we further bring forward a relaxed and regularized DOT model. Meantime, we propose two algorithms to solve the brought forward two models. Finally, extensive experiments on some real datasets are designed in the presence of reflection, large-scale rotation, stretch, noise, and outliers. Some state-of-the-art methods, including CPD, APM, RANSAC, TPS-ICP, TPS-RPM, RPMNet, and classical DOT methods, are to be discussed and compared. For different levels of degradation, the numerical results demonstrate that the proposed methods perform more favorably and robustly than the other methods.

Enhancing VPPA welding quality prediction: A hybrid model integrating prior physical knowledge and CNN analysis

In response to the inconsistency between the features obtained by deep learning models and the quality features reflected by the physical laws of the welding process, this study proposes a solution by integrating a physical prior information model with a CNN model. Initially, the physical laws of the welding process are utilized to annotate the arc, weld pool, and weld seam features relevant to quality, which are then acquired through image processing algorithms, thereby converting the physical laws into a prior information model. Subsequently, this prior information model guides the CNN model for quality recognition, and the CNN model's attention to features is explained through visualization methods to elucidate the relationship between features and quality recognition. Experimental results demonstrate that under the guidance of the prior information model, the CNN model not only automatically focuses on features relevant to quality but also achieves a differential feature attention strategy, thereby improving the recognition accuracy of different outcomes. This research provides a new perspective for deep learning in the field of welding quality recognition.

A Bayesian approach to determine sample size in Erlang single server queues

An essential aspect of the practical application of Erlang single-server queues is the statistical inference of their parameters, notably traffic intensity. This metric is crucial as it serves as a fundamental performance indicator, allowing the derivation of other significant measures, including the mean queue size and the expected number of customers in the system. Additionally, it provides the proportion of time the queue system is occupied. This article explores algorithms for calculating sample sizes for these estimations, based on the number of customers who arrived during service periods, a highly intuitive approach, allowing data collection without concerns about correlations among data points. For this purpose, we have considered two forms of informative priors: Gauss hypergeometric and beta priors. Additionally, a non conjugate and objective prior, known as Jeffreys prior, has been taken into account. We present tables with sample sizes for specific configurations and abacuses for more general configurations obtainable through approximate interpolations.

ELISA-R: an R-based method for robust ELISA data analysis

Enzyme-linked immunosorbent assay (ELISA) is a technique to detect the presence of an antigen or antibody in a sample. ELISA is a simple and cost-effective method that has been used for evaluating vaccine efficacy by detecting the presence of antibodies against viral/bacterial antigens and diagnosis of disease stages. Traditional ELISA data analysis utilizes a standard curve of known analyte, and the concentration of the unknown sample is determined by comparing its observed optical density against the standard curve. However, in the case of vaccine research for complicated bacteria such as Mycobacterium tuberculosis (Mtb), there is no prior information regarding the antigen against which high-affinity antibodies are generated and therefore plotting a standard curve is not feasible. Consequently, the analysis of ELISA data in this instance is based on a comparison between vaccinated and unvaccinated groups. However, to the best of our knowledge, no robust data analysis method exists for “non-standard curve” ELISA. In this paper, we provide a straightforward R-based ELISA data analysis method with open access that incorporates end-point titer determination and curve-fitting models. Our modified method allows for direct measurement data input from the instrument, cleaning and arranging the dataset in the required format, and preparing the final report with calculations while leaving the raw data file unchanged. As an illustration of our method, we provide an example from our published data in which we successfully used our method to compare anti-Mtb antibodies in vaccinated vs non-vaccinated mice.

Tractable optimal experimental design using transport maps

Abstract We present a flexible method for computing Bayesian optimal experimental designs (BOEDs) for inverse problems with intractable posteriors. The approach is applicable to a wide range of BOED problems and can accommodate various optimality criteria, prior distributions and noise models. The key to our approach is the construction of a transport-map-based surrogate to the joint probability law of the design, observational and inference random variables. This order-preserving transport map is constructed using tensor trains and can be used to efficiently sample from (and evaluate approximate densities of) conditional distributions that are required in the evaluation of many commonly-used optimality criteria. The algorithm is also extended to sequential data acquisition problems, where experiments can be performed in sequence to update the state of knowledge about the unknown parameters. The sequential BOED problem is made computationally feasible by preconditioning the approximation of the joint density at the current stage using transport maps constructed at previous stages. The flexibility of our approach in finding optimal designs is illustrated with some numerical examples inspired by disease modeling and the reconstruction of subsurface structures in aquifers.

Kernels of Motor Memory Formation: Temporal Generalization in Bimanual Adaptation.

In daily life, we coordinate both simultaneous and sequential bimanual movements to manipulate objects. Our ability to rapidly account for different object dynamics suggests there are neural mechanisms to quickly deal with them. Here we investigate how actions of one arm can serve as a contextual cue for the other arm, and facilitate adaptation. Specifically, we examine the temporal characteristics that underlie motor memory formation and recall, by testing the contextual effects of prior, simultaneous, and post contralateral arm movements in both male and female human participants. To do so, we measure their temporal generalization in three bimanual interference tasks. Importantly, the timing context of the learned action plays a pivotal role in the temporal generalization. While motor memories trained with post adaptation contextual movements generalize broadly, motor memories trained with prior contextual movements exhibit limited generalization, and motor memories trained with simultaneous contextual movements do not generalize to prior or post contextual timings. This highlights temporal tuning in sensorimotor plasticity: different training conditions yield substantially different temporal generalization characteristics. Since these generalizations extend far beyond any variability in training times, we suggest that the observed differences may stem from inherent differences in the use of prior, current and post-adaptation contextual information in the generation of natural behavior. This would imply differences in the underlying neural circuitry involved in learning and executing the corresponding coordinated bimanual movements.Significance Statement This study addresses a fundamental question in the field of sensorimotor neuroscience of how multiple movements are temporally linked within a single motor memory. We examine the temporal generalization of motor memory formation by varying the timing of contextual movements associated with a learned motor memory across a range of prior, current, and post adaptation movement times using a bimanual motor learning task. We observed distinct patterns of temporal generalization based on whether the contextual movements occurred prior to, simultaneously with, or after the adaptation movement. For the first time, our findings reveal that the timing of contextual movements is crucial in the formation and generalization of motor memories.

Adaptive IMU error correction algorithm for dual-antenna GNSS/IMU integrated vehicle attitude determination

Abstract The attitude of a vehicle can be largely determined by integrating the global navigation satellite system (GNSS) and inertial measurement unit (IMU) in land-based applications. However, traditional dual-antenna GNSS/IMU integrated systems are susceptible to signal reflection, diffraction, and even interruption, leading to reduced accuracy and reliability in challenging environments. To address this issue, this paper proposes an adaptive IMU error correction algorithm for dual-antenna GNSS/IMU integrated vehicle attitude determination. First, we propose an IMU-aided baseline length constraint model to support the ambiguity resolution (AR) process. This model incorporates accurate prior information from the IMU, improving the success rate of AR within the dual-antenna GNSS/IMU integrated system. Second, we present an adaptive IMU error correction mechanism based on long short-term memory (LSTM) and particle swarm optimization (PSO) to assist in vehicle attitude determination and mitigate attitude error drift of the lMU during GNSS outages. Field test results verified that, compared to two other candidate algorithms, the proposed algorithm improves accuracy in roll, pitch, and yaw by 19.23%, 30.56%, and 67.12%, respectively, and by 12.50%, 10.71%, and 38.39% respectively. Moreover, in two distinct scenarios where GNSS is blocked for 120 seconds, the proposed algorithm still provides accurate and stable attitude information for vehicle navigation, with the accuracy of roll, pitch, and yaw kept within 0.08 degrees.