Prior Information Research Articles

Exposure to detectable inaccuracies makes children more diligent fact-checkers of novel claims.

How do children decide when to believe a claim? Here we show that children fact-check claims more and are better able to catch misinformation when they have been exposed to detectable inaccuracies. In two experiments (N = 122), 4-7-year-old children exposed to falsity (as opposed to all true information) sampled more evidence before verifying a test claim in a novel domain. Children's evidentiary standards were graded: fact-checking increased with higher proportions of false statements heard during exposure. A simulation suggests that children's behaviour is adaptive, because increased fact-checking in more dubious environments supports the discovery of potential misinformation. Importantly, children were least diligent at fact-checking a new claim when all prior information was true, suggesting that sanitizing children's informational environments may inadvertently dampen their natural scepticism. Instead, these findings support the counterintuitive possibility that exposing children to some nonsense may scaffold vigilance towards more subtle misinformation in the future.

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.

IRS aided visible light positioning with a single LED transmitter

We propose a visible light positioning (VLP) system with a single light emitting diode (LED) transmitter and an intelligent reflecting surface (IRS) for estimating the position of a receiver equipped with a single photo-detector. By performing a number of transmissions from the LED transmitter and optimizing the orientation vectors of the IRS elements for each transmission, position information is extracted by the receiver based on power measurements of the signals reflecting from the IRS. The theoretical limit and the maximum likelihood (ML) estimator are presented for the proposed setting. In addition, an algorithm, named IRS focusing, is proposed for determining the orientations of the IRS elements during the localization process. The effectiveness of the proposed localization approach is demonstrated through simulations. Furthermore, extensions are provided to apply the proposed approach in the presence of partial prior information about the receiver position and when the IRS is located at the LED transmitter.

Grayscale Iterative Star Spot Extraction Algorithm Based on Image Entropy

Star trackers are susceptible to interference from stray light, such as sunlight, moonlight, and Earth atmosphere light, in the space environment, resulting in an overall improvement in the star image grayscale, poor background uniformity, low star extraction rate, and high number of false star spots. In response to these challenges, this paper proposes a grayscale iterative star spot extraction algorithm based on image entropy. The implementation of the algorithm is mainly divided into two steps: (1) The algorithm conducts multiple grayscale iterations, effectively utilizing the prior information on the local contrast of star spots to filter out stray light backgrounds to a certain extent. (2) By establishing an inner–outer template, the image entropy algorithm is employed to obtain the real star targets to be extracted, which further suppresses the background clutter and noise. Numerical simulations and experimental results demonstrate that, compared to traditional detection algorithms, this algorithm can effectively suppress background stray light, enhance star extraction rates, and reduce the number of false star spots, and it exhibits superior detection performance in complex backgrounds across various scenarios.

Ultra-Short-Term Wind Farm Power Prediction Considering Correlation of Wind Power Fluctuation

Accurate ultra-short-term power prediction for wind farms is challenging under rapid wind speed fluctuations, complicating production planning and power balancing. This paper proposes a new method considering spatial and temporal correlations of wind fluctuations among adjacent wind farms. The method first calculates the time difference between power fluctuations based on wind speed, direction, and relative positions, determining the prior information period. The variational Bayesian model is then used to extract implicit relationships between power fluctuations of adjacent wind farms, enabling power prediction during the prior information period. Finally, the non-prior information period is predicted to complete the ultra-short-term power prediction. Using measured data from three wind farms in Fujian Province, compared to other models, the method demonstrates improved accuracy by effectively leveraging the power fluctuation characteristics of adjacent wind farms, and it has a certain amount of generalizability.

Penalized Bayesian D-Optimal Designs for Regression Models of Continuous Response

We propose extending the use of desirability functions in Bayesian optimal designs for regression models. This technique generates experimental designs with good statistical inference properties according to Bayesian optimal design theory and practical features, as defined by an investigator. These practical features are defined by a penalty function, using an overall desirability function, which is added to a Bayesian $D$-optimal design criterion to penalize impractical experimental designs. This methodology is illustrated by two examples of regression models: quadratic and exponential. Then, we compare designs obtained for different prior distributions of unknown parameters by efficiency calculations and simulation study. Results show that the $D$-efficiencies of the penalized designs relative to the non-penalized Bayesian D-optimal designs are competitive.

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.

Pitfalls in Interpretive Applications of Artificial Intelligence in Radiology.

Interpretive artificial intelligence (AI) tools are poised to change the future of radiology. However, certain pitfalls may pose particular challenges for optimal AI interpretative performance. These include anatomic variants, age-related changes, postoperative changes, medical devices, image artifacts, lack of integration of prior and concurrent imaging examinations and clinical information, and the satisfaction-of-search effect. Model training and development should account for such pitfalls to minimize errors and optimize interpretation accuracy. More broadly, AI algorithms should be exposed to diverse and complex training datasets to yield a holistic interpretation that considers all relevant information beyond the individual examination. Successful clinical deployment of AI tools will require that radiologist end users recognize these pitfalls and other limitations of the available models. Furthermore, developers should incorporate explainable AI techniques (e.g., heat maps) into their tools, to improve radiologists' understanding of model outputs and to enable radiologists to provide feedback for guiding continuous learning and iterative refinement. In this article, we provide an overview of common pitfalls that radiologists may encounter when using interpretive AI products in daily practice. We present how such pitfalls lead to AI errors and offer potential strategies that AI developers may use for their mitigation.

Dual structure-aware image filterings for semi-supervised medical image segmentation

Semi-supervised image segmentation has attracted great attention recently. The key is how to leverage unlabeled images in the training process. Most methods maintain consistent predictions of the unlabeled images under variations (e.g., adding noise/perturbations, or creating alternative versions) in the image and/or model level. In most image-level variation, medical images often have prior structure information, which has not been well explored. In this paper, we propose novel dual structure-aware image filterings (DSAIF) as the image-level variations for semi-supervised medical image segmentation. Motivated by connected filtering that simplifies image via filtering in structure-aware tree-based image representation, we resort to the dual contrast invariant Max-tree and Min-tree representation. Specifically, we propose a novel connected filtering that removes topologically equivalent nodes (i.e. connected components) having no siblings in the Max/Min-tree. This results in two filtered images preserving topologically critical structure. Applying the proposed DSAIF to mutually supervised networks decreases the consensus of their erroneous predictions on unlabeled images. This helps to alleviate the confirmation bias issue of overfitting to noisy pseudo labels of unlabeled images, and thus effectively improves the segmentation performance. Extensive experimental results on three benchmark datasets demonstrate that the proposed method significantly/consistently outperforms some state-of-the-art methods. The source codes will be publicly available.

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.

Ultrafast metaproteomics for quantitative assessment of strain isolates and microbiomes

BackgroundMicrobial communities are essential in human health and environmental regulation, but present a challenge for the analytical science due to their diversity and dynamic range. Tandem mass spectrometry provides functional insights on microbial life cycle, but is time-consuming. MALDI TOF excels in rapid species identification, but not functional assessment. To address critical challenges in human health and environmental sustainability, microbiology needs advanced mass spectrometry methods and bioinformatic tools enabling both rapid identification and accurate assessment of functional activity of microbial communities. ResultsWe show for the first time that both identity and functional activity of microorganisms and their communities can be accurately determined in experiments as short as 7 min per sample, using the basic Orbitrap MS configuration without peptide fragmentation. The approach was validated using strain isolates, mock microbiomes composed of bacteria spiked at known concentrations and human fecal microbiomes. Our new bioinformatic algorithm identifies the bacterial species with an accuracy of 95 %, when no prior information on the sample is available. Microbiome composition was resolved at the genus level with the mean difference between the actual and identified components of 12 %. For mock microbiomes, Pearson coefficient of up to 0.97 was achieved in estimates of strain biomass change. By the example of Rhodococcus biodegradation of n-alkanes, phenols and its derivatives, we showed the accurate assessment of functional activity of strain isolates, compared with the standard label-free and label-based approaches. SignificanceOur approach makes microbial proteomics fast, functional and insightful using the Orbitrap instruments even without employing peptide fragmentation technology. The approach can be applied to any microorganisms and can take a niche in routine functional assessment of microbial pathogens and consortiums in clinical diagnostics together with MALDI TOF MS and 16S rRNA gene sequencing.

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%.