The integrated nested Laplace approximation(INLA)algorithm provides a computationally efficient approach for approximate Bayesian inference,overcoming the limitations of traditional Markov chain Monte Carlo(MCMC)metho...The integrated nested Laplace approximation(INLA)algorithm provides a computationally efficient approach for approximate Bayesian inference,overcoming the limitations of traditional Markov chain Monte Carlo(MCMC)methods.This paper reviews INLA algorithm and provides a systematic review of six key books that explore the theoretical foundations,practical implementations,and diverse applications of INLA.These six books cover spatial and spatio-temporal modelling,general Bayesian inference,SPDE-based spatial analysis,geospatial health data,regression modelling,and dynamic time series.In addition,these books highlight the versatility of INLA method in handling complex models while maintaining high computational efficiency.This paper begins with an introduction to the INLA method and algorithm,followed by a systematic review of six key publications in the field.展开更多
Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-h...Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-harm,long-term disability,reduced productivity,and significant societal and economic burden.Despite recent advances,detecting risk from online text remains challenging due to heterogeneous language,evolving semantics,and the sequential emergence of new datasets.Effective solutions must encode clinically meaningful cues,reason about causal relations,and adapt to new domains without forgetting prior knowledge.To address these challenges,this paper presents a Continual Neuro-Symbolic Graph Learning(CNSGL)framework that unifies symbolic reasoning,causal inference,and continual learning within a single architecture.Each post is represented as a symbolic graph linking clinically relevant tags to textual content,enriched with causal edges derived from directional Point-wise Mutual Information(PMI).A two-layer Graph Convolutional Network(GCN)encodes these graphs,and a Transformer-based attention pooler aggregates node embeddings while providing interpretable tag-level importances.Continual adaptation across datasets is achieved through the Multi-Head Freeze(MH-Freeze)strategy,which freezes a shared encoder and incrementally trains lightweight task-specific heads(small classifiers attached to the shared embedding).Experimental evaluations across six diverse mental-health datasets ranging from Reddit discourse to clinical interviews,demonstrate that MH-Freeze consistently outperforms existing continual-learning baselines in both discriminative accuracy and calibration reliability.Across six datasets,MH-Freeze achieves up to 0.925 accuracy and 0.923 F1-Score,with AUPRC≥0.934 and AUROC≥0.942,consistently surpassing all continual-learning baselines.The results confirm the framework’s ability to preserve prior knowledge,adapt to domain shifts,and maintain causal interpretability,establishing CNSGL as a promising step toward robust,explainable,and lifelong mental-health risk assessment.展开更多
Transformer models face significant computational challenges in private inference(PI).Existing optimization methods often rely on isolated techniques,neglecting joint structural and operational improvements.We propose...Transformer models face significant computational challenges in private inference(PI).Existing optimization methods often rely on isolated techniques,neglecting joint structural and operational improvements.We propose IG-3D,a unified framework that integrates structured compression and operator approximation through accurate importance assessment.Our approach first evaluates attention head importance using Integrated Gradients(IG),offering greater stability and theoretical soundness than gradient-based methods.We then apply a threedimensional optimization:(1)structurally pruning redundant attention heads;(2)replacing Softmax with adaptive polynomial approximation to avoid exponential computations;(3)implementing layer-wise GELU substitution to accommodate different layer characteristics.A joint thresholdmechanism coordinates compression across dimensions under accuracy constraints.Experimental results on the GLUE benchmark show that our method achieves an average 2.9×speedup in inference latency and a 50%reduction in communication cost,while controlling the accuracy loss within 2.3%,demonstrating significant synergistic effects and a superior accuracy-efficiency trade-off compared to single-technique optimization strategies.展开更多
The detection of gravitational waves by the LIGO-Virgo-KAGRA collaboration has ushered in a new era of observational astronomy,emphasizing the need for rapid and detailed parameter estimation and population-level anal...The detection of gravitational waves by the LIGO-Virgo-KAGRA collaboration has ushered in a new era of observational astronomy,emphasizing the need for rapid and detailed parameter estimation and population-level analyses.Traditional Bayesian inference methods,particularly Markov chain Monte Carlo,face significant computational challenges when dealing with the high-dimensional parameter spaces and complex noise characteristics inherent in gravitational wave data.This review examines the emerging role of simulation-based inference methods in gravitational wave astronomy,with a focus on approaches that leverage machine-learning techniques such as normalizing flows and neural posterior estimation.We provide a comprehensive overview of the theoretical foundations underlying various simulation-based inference methods,including neural posterior estimation,neural ratio estimation,neural likelihood estimation,flow matching,and consistency models.We explore the applications of these methods across diverse gravitational wave data processing scenarios,from single-source parameter estimation and overlapping signal analysis to testing general relativity and conducting population studies.Although these techniques demonstrate speed improvements over traditional methods in controlled studies,their model-dependent nature and sensitivity to prior assumptions are barriers to their widespread adoption.Their accuracy,which is similar to that of conventional methods,requires further validation across broader parameter spaces and noise conditions.展开更多
1.Introduction Data inference(DInf)is a data security threat in which critical information is inferred from low-sensitivity data.Once regarded as an advanced professional threat limited to intelligence analysts,DInf h...1.Introduction Data inference(DInf)is a data security threat in which critical information is inferred from low-sensitivity data.Once regarded as an advanced professional threat limited to intelligence analysts,DInf has become a widespread risk in the artificial intelligence(AI)era.展开更多
In 6G,artificial intelligence represented by deep nerual network(DNN)will unleash its potential and empower IoT applications to transform into intelligent IoT applications.However,whole DNNbased inference is difficult...In 6G,artificial intelligence represented by deep nerual network(DNN)will unleash its potential and empower IoT applications to transform into intelligent IoT applications.However,whole DNNbased inference is difficult to carry out on resourceconstrained intelligent IoT devices and will suffer privacy leakage when offloading to the cloud or mobile edge computation server(MECs).In this paper,we formulate a privacy and delay dual-driven device-edge collaborative inference(P4DE-CI)system to preserve the privacy of raw data while accelerating the intelligent inference process,where the intelligent IoT devices run the front-end part of DNN model and the MECs execute the back-end part of DNN model.Considering three typical privacy leakage models and the end-to-end delay of collaborative DNN-based inference,we define a novel intelligent inference Quality of service(I2-QoS)metric as the weighted summation of the inference latency and privacy preservation level.Moreover,we propose a DDPG-based joint DNN model optimization and resource allocation algorithm to maximize I2-QoS,by optimizing the association relationship between intelligent IoT devices and MECs,the DNN model placement decision,and the DNN model partition decision.Experiments carried out on the AlexNet model reveal that the proposed algorithm has better performance in both privacy-preserving and inference-acceleration.展开更多
Understanding how renewable energy generation affects electricity prices is essential for designing efficient and sustainable electricity markets.However,most existing studies rely on regression-based approaches that ...Understanding how renewable energy generation affects electricity prices is essential for designing efficient and sustainable electricity markets.However,most existing studies rely on regression-based approaches that capture correlations but fail to identify causal relationships,particularly in the presence of non-linearities and confounding factors.This limits their value for informing policy and market design in the context of the energy transition.To address this gap,we propose a novel causal inference framework based on local partially linear double machine learning(DML).Our method isolates the true impact of predicted wind and solar power generation on electricity prices by controlling for high-dimensional confounders and allowing for non-linear,context-dependent effects.This represents a substantial methodological advancement over standard econometric techniques.Applying this framework to the UK electricity market over the period 2018-2024,we produce the first robust causal estimates of how renewables affect dayahead wholesale electricity prices.We find that wind power exerts a U-shaped causal effect:at low penetration levels,a 1 GWh increase reduces prices by up to£7/MWh,the effect weakens at mid-levels,and intensifies again at higher penetration.Solar power consistently reduces prices at low penetration levels,up to£9/MWh per additional GWh,but its marginal effect diminishes quickly.Importantly,the magnitude of these effects has increased over time,reflecting the growing influence of renewables on price formation as their share in the energy mix rises.These findings offer a sound empirical basis for improving the design of support schemes,refining capacity planning,and enhancing electricity market efficiency.By providing a robust causal understanding of renewable impacts,our study contributes both methodological innovation and actionable insights to guide future energy policy.展开更多
Robustness against measurement uncertainties is crucial for gas turbine engine diagnosis.While current research focuses mainly on measurement noise,measurement bias remains challenging.This study proposes a novel perf...Robustness against measurement uncertainties is crucial for gas turbine engine diagnosis.While current research focuses mainly on measurement noise,measurement bias remains challenging.This study proposes a novel performance-based fault detection and identification(FDI)strategy for twin-shaft turbofan gas turbine engines and addresses these uncertainties through a first-order Takagi-Sugeno-Kang fuzzy inference system.To handle ambient condition changes,we use parameter correction to preprocess the raw measurement data,which reduces the FDI’s system complexity.Additionally,the power-level angle is set as a scheduling parameter to reduce the number of rules in the TSK-based FDI system.The data for designing,training,and testing the proposed FDI strategy are generated using a component-level turbofan engine model.The antecedent and consequent parameters of the TSK-based FDI system are optimized using the particle swarm optimization algorithm and ridge regression.A robust structure combining a specialized fuzzy inference system with the TSK-based FDI system is proposed to handle measurement biases.The performance of the first-order TSK-based FDI system and robust FDI structure are evaluated through comprehensive simulation studies.Comparative studies confirm the superior accuracy of the first-order TSK-based FDI system in fault detection,isolation,and identification.The robust structure demonstrates a 2%-8%improvement in the success rate index under relatively large measurement bias conditions,thereby indicating excellent robustness.Accuracy against significant bias values and computation time are also evaluated,suggesting that the proposed robust structure has desirable online performance.This study proposes a novel FDI strategy that effectively addresses measurement uncertainties.展开更多
Edge Machine Learning(EdgeML)and Tiny Machine Learning(TinyML)are fast-growing fields that bring machine learning to resource-constrained devices,allowing real-time data processing and decision-making at the network’...Edge Machine Learning(EdgeML)and Tiny Machine Learning(TinyML)are fast-growing fields that bring machine learning to resource-constrained devices,allowing real-time data processing and decision-making at the network’s edge.However,the complexity of model conversion techniques,diverse inference mechanisms,and varied learning strategies make designing and deploying these models challenging.Additionally,deploying TinyML models on resource-constrained hardware with specific software frameworks has broadened EdgeML’s applications across various sectors.These factors underscore the necessity for a comprehensive literature review,as current reviews do not systematically encompass the most recent findings on these topics.Consequently,it provides a comprehensive overview of state-of-the-art techniques in model conversion,inference mechanisms,learning strategies within EdgeML,and deploying these models on resource-constrained edge devices using TinyML.It identifies 90 research articles published between 2018 and 2025,categorizing them into two main areas:(1)model conversion,inference,and learning strategies in EdgeML and(2)deploying TinyML models on resource-constrained hardware using specific software frameworks.In the first category,the synthesis of selected research articles compares and critically reviews various model conversion techniques,inference mechanisms,and learning strategies.In the second category,the synthesis identifies and elaborates on major development boards,software frameworks,sensors,and algorithms used in various applications across six major sectors.As a result,this article provides valuable insights for researchers,practitioners,and developers.It assists them in choosing suitable model conversion techniques,inference mechanisms,learning strategies,hardware development boards,software frameworks,sensors,and algorithms tailored to their specific needs and applications across various sectors.展开更多
Developing efficient neural network(NN)computing systems is crucial in the era of artificial intelligence(AI).Traditional von Neumann architectures have both the issues of"memory wall"and"power wall&quo...Developing efficient neural network(NN)computing systems is crucial in the era of artificial intelligence(AI).Traditional von Neumann architectures have both the issues of"memory wall"and"power wall",limiting the data transfer between memory and processing units[1,2].Compute-in-memory(CIM)technologies,particularly analogue CIM with memristor crossbars,are promising because of their high energy efficiency,computational parallelism,and integration density for NN computations[3].In practical applications,analogue CIM excels in tasks like speech recognition and image classification,revealing its unique advantages.For instance,it efficiently processes vast amounts of audio data in speech recognition,achieving high accuracy with minimal power consumption.In image classification,the high parallelism of analogue CIM significantly speeds up feature extraction and reduces processing time.With the boosting development of AI applications,the demands for computational accuracy and task complexity are rising continually.However,analogue CIM systems are limited in handling complex regression tasks with needs of precise floating-point(FP)calculations.They are primarily suited for the classification tasks with low data precision and a limited dynamic range[4].展开更多
Published proof test coverage(PTC)estimates for emergency shutdown valves(ESDVs)show only moderate agreement and are predominantly opinion-based.A Failure Modes,Effects,and Diagnostics Analysis(FMEDA)was undertaken us...Published proof test coverage(PTC)estimates for emergency shutdown valves(ESDVs)show only moderate agreement and are predominantly opinion-based.A Failure Modes,Effects,and Diagnostics Analysis(FMEDA)was undertaken using component failure rate data to predict PTC for a full stroke test and a partial stroke test.Given the subjective and uncertain aspects of the FMEDA approach,specifically the selection of component failure rates and the determination of the probability of detecting failure modes,a Fuzzy Inference System(FIS)was proposed to manage the data,addressing the inherent uncertainties.Fuzzy inference systems have been used previously for various FMEA type assessments,but this is the first time an FIS has been employed for use with FMEDA.ESDV PTC values were generated from both the standard FMEDA and the fuzzy-FMEDA approaches using data provided by FMEDA experts.This work demonstrates that fuzzy inference systems can address the subjectivity inherent in FMEDA data,enabling reliable estimates of ESDV proof test coverage for both full and partial stroke tests.This facilitates optimized maintenance planning while ensuring safety is not compromised.展开更多
In order to solve the problems of high experimental cost of ammunition,lack of field test data,and the difficulty in applying the ammunition hit probability estimation method in classical statistics,this paper assumes...In order to solve the problems of high experimental cost of ammunition,lack of field test data,and the difficulty in applying the ammunition hit probability estimation method in classical statistics,this paper assumes that the projectile dispersion of ammunition is a two-dimensional joint normal distribution,and proposes a new Bayesian inference method of ammunition hit probability based on normal-inverse Wishart distribution.Firstly,the conjugate joint prior distribution of the projectile dispersion characteristic parameters is determined to be a normal inverse Wishart distribution,and the hyperparameters in the prior distribution are estimated by simulation experimental data and historical measured data.Secondly,the field test data is integrated with the Bayesian formula to obtain the joint posterior distribution of the projectile dispersion characteristic parameters,and then the hit probability of the ammunition is estimated.Finally,compared with the binomial distribution method,the method in this paper can consider the dispersion information of ammunition projectiles,and the hit probability information is more fully utilized.The hit probability results are closer to the field shooting test samples.This method has strong applicability and is conducive to obtaining more accurate hit probability estimation results.展开更多
Federated Graph Neural Networks (FedGNNs) have achieved significant success in representation learning for graph data, enabling collaborative training among multiple parties without sharing their raw graph data and so...Federated Graph Neural Networks (FedGNNs) have achieved significant success in representation learning for graph data, enabling collaborative training among multiple parties without sharing their raw graph data and solving the data isolation problem faced by centralized GNNs in data-sensitive scenarios. Despite the plethora of prior work on inference attacks against centralized GNNs, the vulnerability of FedGNNs to inference attacks has not yet been widely explored. It is still unclear whether the privacy leakage risks of centralized GNNs will also be introduced in FedGNNs. To bridge this gap, we present PIAFGNN, the first property inference attack (PIA) against FedGNNs. Compared with prior works on centralized GNNs, in PIAFGNN, the attacker can only obtain the global embedding gradient distributed by the central server. The attacker converts the task of stealing the target user’s local embeddings into a regression problem, using a regression model to generate the target graph node embeddings. By training shadow models and property classifiers, the attacker can infer the basic property information within the target graph that is of interest. Experiments on three benchmark graph datasets demonstrate that PIAFGNN achieves attack accuracy of over 70% in most cases, even approaching the attack accuracy of inference attacks against centralized GNNs in some instances, which is much higher than the attack accuracy of the random guessing method. Furthermore, we observe that common defense mechanisms cannot mitigate our attack without affecting the model’s performance on mainly classification tasks.展开更多
Osteoporosis is a known risk factor for rotator cuff tears(RCTs),but the causal correlation and underlying mechanisms remain unclear.This study aims to evaluate the impact of osteoporosis on RCT risk and investigate t...Osteoporosis is a known risk factor for rotator cuff tears(RCTs),but the causal correlation and underlying mechanisms remain unclear.This study aims to evaluate the impact of osteoporosis on RCT risk and investigate their genetic associations.Using data from the UK Biobank(n=457871),cross-sectional analyses demonstrated that osteoporosis was significantly associated with an increased risk of RCTs(adjusted OR[95%CI]=1.38[1.25–1.52]).A longitudinal analysis of a subset of patients(n=268117)over 11 years revealed that osteoporosis increased the risk of RCTs(adjusted HR[95%CI]=1.56[1.29–1.87]),which is notably varied between sexes in sex-stratified analysis.Causal inference methods,including propensity score matching,inverse probability weighting,causal random forest and survival random forest models further confirmed the causal effect,both from cross-sectional and longitudinal perspectives.展开更多
Offshore drilling costs are high,and the downhole environment is even more complex.Improving the rate of penetration(ROP)can effectively shorten offshore drilling cycles and improve economic benefits.It is difficult f...Offshore drilling costs are high,and the downhole environment is even more complex.Improving the rate of penetration(ROP)can effectively shorten offshore drilling cycles and improve economic benefits.It is difficult for the current ROP models to guarantee the prediction accuracy and the robustness of the models at the same time.To address the current issues,a new ROP prediction model was developed in this study,which considers ROP as a time series signal(ROP signal).The model is based on the time convolutional network(TCN)framework and integrates ensemble empirical modal decomposition(EEMD)and Bayesian network causal inference(BN),the model is named EEMD-BN-TCN.Within the proposed model,the EEMD decomposes the original ROP signal into multiple sets of sub-signals.The BN determines the causal relationship between the sub-signals and the key physical parameters(weight on bit and revolutions per minute)and carries out preliminary reconstruction of the sub-signals based on the causal relationship.The TCN predicts signals reconstructed by BN.When applying this model to an actual production well,the average absolute percentage error of the EEMD-BN-TCN prediction decreased from 18.4%with TCN to 9.2%.In addition,compared with other models,the EEMD-BN-TCN can improve the decomposition signal of ROP by regulating weight on bit and revolutions per minute,ultimately enhancing ROP.展开更多
Inferring phylogenetic trees from molecular sequences is a cornerstone of evolutionary biology.Many standard phylogenetic methods(such as maximum-likelihood[ML])rely on explicit models of sequence evolution and thus o...Inferring phylogenetic trees from molecular sequences is a cornerstone of evolutionary biology.Many standard phylogenetic methods(such as maximum-likelihood[ML])rely on explicit models of sequence evolution and thus often suffer from model misspecification or inadequacy.The on-rising deep learning(DL)techniques offer a powerful alternative.Deep learning employs multi-layered artificial neural networks to progressively transform input data into more abstract and complex representations.DL methods can autonomously uncover meaningful patterns from data,thereby bypassing potential biases introduced by predefined features(Franklin,2005;Murphy,2012).Recent efforts have aimed to apply deep neural networks(DNNs)to phylogenetics,with a growing number of applications in tree reconstruction(Suvorov et al.,2020;Zou et al.,2020;Nesterenko et al.,2022;Smith and Hahn,2023;Wang et al.,2023),substitution model selection(Abadi et al.,2020;Burgstaller-Muehlbacher et al.,2023),and diversification rate inference(Voznica et al.,2022;Lajaaiti et al.,2023;Lambert et al.,2023).In phylogenetic tree reconstruction,PhyDL(Zou et al.,2020)and Tree_learning(Suvorov et al.,2020)are two notable DNN-based programs designed to infer unrooted quartet trees directly from alignments of four amino acid(AA)and DNA sequences,respectively.展开更多
Modern compression and acceleration methods for exploring efficient deep neural networks render real-world applications more feasible.Existing approaches uniformly apply the same procedure to every input image,overloo...Modern compression and acceleration methods for exploring efficient deep neural networks render real-world applications more feasible.Existing approaches uniformly apply the same procedure to every input image,overlooking instancewise complexity variations.Moreover,owing to pruning or decomposition techniques,the upper bound of network representation capabilities might be permanently diminished.In this work,an input-dependent multiscale dynamic inference method(MSDI)is developed to strike a better balance between model performance and inference acceleration.Specifically,we modify the main body of a convolutional network to obtain a series of parameter-sharing subnetworks with varying levels of complexity.A side branch structure is then introduced to assign an input instance to a suitable subnetwork as its inference route,and we expect to accelerate the inference by assigning the easy input to the subnetwork with low capacity.We further propose multiscale distillation training to optimize the training of the modified subnetworks.Additionally,we compare the entropy-based and learning-based grading approaches,aiming to obtain a more suitable route assignment method.Experiments show that MSDI can accelerate most existing convolutional models,achieving up to 74.7%computation savings across diverse datasets.展开更多
Protocol Reverse Engineering(PRE)is of great practical importance in Internet security-related fields such as intrusion detection,vulnerability mining,and protocol fuzzing.For unknown binary protocols having fixed-len...Protocol Reverse Engineering(PRE)is of great practical importance in Internet security-related fields such as intrusion detection,vulnerability mining,and protocol fuzzing.For unknown binary protocols having fixed-length fields,and the accurate identification of field boundaries has a great impact on the subsequent analysis and final performance.Hence,this paper proposes a new protocol segmentation method based on Information-theoretic statistical analysis for binary protocols by formulating the field segmentation of unsupervised binary protocols as a probabilistic inference problem and modeling its uncertainty.Specifically,we design four related constructions between entropy changes and protocol field segmentation,introduce random variables,and construct joint probability distributions with traffic sample observations.Probabilistic inference is then performed to identify the possible protocol segmentation points.Extensive trials on nine common public and industrial control protocols show that the proposed method yields higher-quality protocol segmentation results.展开更多
Chronic diseases are major causes of global death and disability,significantly impacting individual health and imposing economic burdens.This study aims to investigate the causal relationship between physical activity...Chronic diseases are major causes of global death and disability,significantly impacting individual health and imposing economic burdens.This study aims to investigate the causal relationship between physical activity(PA)and the development of chronic diseases.Using a Mendelian randomization(MR)approach,we incorporated average PA and its subtypes(more than 450000 participants)as exposure measures and eight chronic diseases as outcome measures.Data were obtained from the European Genome-Wide Association Study(GWAS).The primary causal analysis technique employed was the inverse variance weighting(IVW)method,with MR-Egger,weighted median,weighted mode,and simple mode used to validate the results.Sensitivity analyses were performed to assess heterogeneity and pleiotropy.The IVW approach results show that vigorous physical activity(VPA)is associated with a modest reduction in the risk of major coronary heart disease(OR=0.95,95%CI:0.91–0.99,p=0.01).The causal directions of the other four MR methods are consistent with this result and validated by sensitivity analysis.No substantial associations were found between other levels of PA and chronic disease.Our findings underscore the importance of VPA in preventive cardiology and suggest its potential role in public health initiatives.Further research should explore the impact of PA on different demographic groups and the doseresponse relationship of VPA on heart health.展开更多
Future 6G communications will open up opportunities for innovative applications,including Cyber-Physical Systems,edge computing,supporting Industry 5.0,and digital agriculture.While automation is creating efficiencies...Future 6G communications will open up opportunities for innovative applications,including Cyber-Physical Systems,edge computing,supporting Industry 5.0,and digital agriculture.While automation is creating efficiencies,it can also create new cyber threats,such as vulnerabilities in trust and malicious node injection.Denialof-Service(DoS)attacks can stop many forms of operations by overwhelming networks and systems with data noise.Current anomaly detection methods require extensive software changes and only detect static threats.Data collection is important for being accurate,but it is often a slow,tedious,and sometimes inefficient process.This paper proposes a new wavelet transformassisted Bayesian deep learning based probabilistic(WT-BDLP)approach tomitigate malicious data injection attacks in 6G edge networks.The proposed approach combines outlier detection based on a Bayesian learning conditional variational autoencoder(Bay-LCVariAE)and traffic pattern analysis based on continuous wavelet transform(CWT).The Bay-LCVariAE framework allows for probabilistic modelling of generative features to facilitate capturing how features of interest change over time,spatially,and for recognition of anomalies.Similarly,CWT allows emphasizing the multi-resolution spectral analysis and permits temporally relevant frequency pattern recognition.Experimental testing showed that the flexibility of the Bayesian probabilistic framework offers a vast improvement in anomaly detection accuracy over existing methods,with a maximum accuracy of 98.21%recognizing anomalies.展开更多
基金supported by the National Natural Science Foundation of China[grant number 12001266]the Humanities and Social Science Projects ofMinistry of Education of China[grant number 19YJCZH166]supported by the National Natural Science Foundation of China[grant numbers 12271168 and 12531013].
文摘The integrated nested Laplace approximation(INLA)algorithm provides a computationally efficient approach for approximate Bayesian inference,overcoming the limitations of traditional Markov chain Monte Carlo(MCMC)methods.This paper reviews INLA algorithm and provides a systematic review of six key books that explore the theoretical foundations,practical implementations,and diverse applications of INLA.These six books cover spatial and spatio-temporal modelling,general Bayesian inference,SPDE-based spatial analysis,geospatial health data,regression modelling,and dynamic time series.In addition,these books highlight the versatility of INLA method in handling complex models while maintaining high computational efficiency.This paper begins with an introduction to the INLA method and algorithm,followed by a systematic review of six key publications in the field.
基金supported by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(RS-2025-00518960)in part by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(RS-2025-00563192).
文摘Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-harm,long-term disability,reduced productivity,and significant societal and economic burden.Despite recent advances,detecting risk from online text remains challenging due to heterogeneous language,evolving semantics,and the sequential emergence of new datasets.Effective solutions must encode clinically meaningful cues,reason about causal relations,and adapt to new domains without forgetting prior knowledge.To address these challenges,this paper presents a Continual Neuro-Symbolic Graph Learning(CNSGL)framework that unifies symbolic reasoning,causal inference,and continual learning within a single architecture.Each post is represented as a symbolic graph linking clinically relevant tags to textual content,enriched with causal edges derived from directional Point-wise Mutual Information(PMI).A two-layer Graph Convolutional Network(GCN)encodes these graphs,and a Transformer-based attention pooler aggregates node embeddings while providing interpretable tag-level importances.Continual adaptation across datasets is achieved through the Multi-Head Freeze(MH-Freeze)strategy,which freezes a shared encoder and incrementally trains lightweight task-specific heads(small classifiers attached to the shared embedding).Experimental evaluations across six diverse mental-health datasets ranging from Reddit discourse to clinical interviews,demonstrate that MH-Freeze consistently outperforms existing continual-learning baselines in both discriminative accuracy and calibration reliability.Across six datasets,MH-Freeze achieves up to 0.925 accuracy and 0.923 F1-Score,with AUPRC≥0.934 and AUROC≥0.942,consistently surpassing all continual-learning baselines.The results confirm the framework’s ability to preserve prior knowledge,adapt to domain shifts,and maintain causal interpretability,establishing CNSGL as a promising step toward robust,explainable,and lifelong mental-health risk assessment.
文摘Transformer models face significant computational challenges in private inference(PI).Existing optimization methods often rely on isolated techniques,neglecting joint structural and operational improvements.We propose IG-3D,a unified framework that integrates structured compression and operator approximation through accurate importance assessment.Our approach first evaluates attention head importance using Integrated Gradients(IG),offering greater stability and theoretical soundness than gradient-based methods.We then apply a threedimensional optimization:(1)structurally pruning redundant attention heads;(2)replacing Softmax with adaptive polynomial approximation to avoid exponential computations;(3)implementing layer-wise GELU substitution to accommodate different layer characteristics.A joint thresholdmechanism coordinates compression across dimensions under accuracy constraints.Experimental results on the GLUE benchmark show that our method achieves an average 2.9×speedup in inference latency and a 50%reduction in communication cost,while controlling the accuracy loss within 2.3%,demonstrating significant synergistic effects and a superior accuracy-efficiency trade-off compared to single-technique optimization strategies.
基金supported by the National Key Research and Development Program of China(2021YFC2203004)the National Natural Science Foundation of China(NSFC)(12405076,12247187,and 12147103)+1 种基金the National Astronomical Data Center(NADC2023YDS-01)the Fundamental Research Funds for the Central Universities.
文摘The detection of gravitational waves by the LIGO-Virgo-KAGRA collaboration has ushered in a new era of observational astronomy,emphasizing the need for rapid and detailed parameter estimation and population-level analyses.Traditional Bayesian inference methods,particularly Markov chain Monte Carlo,face significant computational challenges when dealing with the high-dimensional parameter spaces and complex noise characteristics inherent in gravitational wave data.This review examines the emerging role of simulation-based inference methods in gravitational wave astronomy,with a focus on approaches that leverage machine-learning techniques such as normalizing flows and neural posterior estimation.We provide a comprehensive overview of the theoretical foundations underlying various simulation-based inference methods,including neural posterior estimation,neural ratio estimation,neural likelihood estimation,flow matching,and consistency models.We explore the applications of these methods across diverse gravitational wave data processing scenarios,from single-source parameter estimation and overlapping signal analysis to testing general relativity and conducting population studies.Although these techniques demonstrate speed improvements over traditional methods in controlled studies,their model-dependent nature and sensitivity to prior assumptions are barriers to their widespread adoption.Their accuracy,which is similar to that of conventional methods,requires further validation across broader parameter spaces and noise conditions.
基金supported by the National Key Research and Development Program of China(2022YFB2703503)the National Natural Science Foundation of China(62293501,62525210,and 62293502)the China Scholarship Council(202306280318).
文摘1.Introduction Data inference(DInf)is a data security threat in which critical information is inferred from low-sensitivity data.Once regarded as an advanced professional threat limited to intelligence analysts,DInf has become a widespread risk in the artificial intelligence(AI)era.
基金supported by the National Natural Science Foundation of China(No.62201079)the Beijing Natural Science Foundation(No.L232051).
文摘In 6G,artificial intelligence represented by deep nerual network(DNN)will unleash its potential and empower IoT applications to transform into intelligent IoT applications.However,whole DNNbased inference is difficult to carry out on resourceconstrained intelligent IoT devices and will suffer privacy leakage when offloading to the cloud or mobile edge computation server(MECs).In this paper,we formulate a privacy and delay dual-driven device-edge collaborative inference(P4DE-CI)system to preserve the privacy of raw data while accelerating the intelligent inference process,where the intelligent IoT devices run the front-end part of DNN model and the MECs execute the back-end part of DNN model.Considering three typical privacy leakage models and the end-to-end delay of collaborative DNN-based inference,we define a novel intelligent inference Quality of service(I2-QoS)metric as the weighted summation of the inference latency and privacy preservation level.Moreover,we propose a DDPG-based joint DNN model optimization and resource allocation algorithm to maximize I2-QoS,by optimizing the association relationship between intelligent IoT devices and MECs,the DNN model placement decision,and the DNN model partition decision.Experiments carried out on the AlexNet model reveal that the proposed algorithm has better performance in both privacy-preserving and inference-acceleration.
文摘Understanding how renewable energy generation affects electricity prices is essential for designing efficient and sustainable electricity markets.However,most existing studies rely on regression-based approaches that capture correlations but fail to identify causal relationships,particularly in the presence of non-linearities and confounding factors.This limits their value for informing policy and market design in the context of the energy transition.To address this gap,we propose a novel causal inference framework based on local partially linear double machine learning(DML).Our method isolates the true impact of predicted wind and solar power generation on electricity prices by controlling for high-dimensional confounders and allowing for non-linear,context-dependent effects.This represents a substantial methodological advancement over standard econometric techniques.Applying this framework to the UK electricity market over the period 2018-2024,we produce the first robust causal estimates of how renewables affect dayahead wholesale electricity prices.We find that wind power exerts a U-shaped causal effect:at low penetration levels,a 1 GWh increase reduces prices by up to£7/MWh,the effect weakens at mid-levels,and intensifies again at higher penetration.Solar power consistently reduces prices at low penetration levels,up to£9/MWh per additional GWh,but its marginal effect diminishes quickly.Importantly,the magnitude of these effects has increased over time,reflecting the growing influence of renewables on price formation as their share in the energy mix rises.These findings offer a sound empirical basis for improving the design of support schemes,refining capacity planning,and enhancing electricity market efficiency.By providing a robust causal understanding of renewable impacts,our study contributes both methodological innovation and actionable insights to guide future energy policy.
文摘Robustness against measurement uncertainties is crucial for gas turbine engine diagnosis.While current research focuses mainly on measurement noise,measurement bias remains challenging.This study proposes a novel performance-based fault detection and identification(FDI)strategy for twin-shaft turbofan gas turbine engines and addresses these uncertainties through a first-order Takagi-Sugeno-Kang fuzzy inference system.To handle ambient condition changes,we use parameter correction to preprocess the raw measurement data,which reduces the FDI’s system complexity.Additionally,the power-level angle is set as a scheduling parameter to reduce the number of rules in the TSK-based FDI system.The data for designing,training,and testing the proposed FDI strategy are generated using a component-level turbofan engine model.The antecedent and consequent parameters of the TSK-based FDI system are optimized using the particle swarm optimization algorithm and ridge regression.A robust structure combining a specialized fuzzy inference system with the TSK-based FDI system is proposed to handle measurement biases.The performance of the first-order TSK-based FDI system and robust FDI structure are evaluated through comprehensive simulation studies.Comparative studies confirm the superior accuracy of the first-order TSK-based FDI system in fault detection,isolation,and identification.The robust structure demonstrates a 2%-8%improvement in the success rate index under relatively large measurement bias conditions,thereby indicating excellent robustness.Accuracy against significant bias values and computation time are also evaluated,suggesting that the proposed robust structure has desirable online performance.This study proposes a novel FDI strategy that effectively addresses measurement uncertainties.
文摘Edge Machine Learning(EdgeML)and Tiny Machine Learning(TinyML)are fast-growing fields that bring machine learning to resource-constrained devices,allowing real-time data processing and decision-making at the network’s edge.However,the complexity of model conversion techniques,diverse inference mechanisms,and varied learning strategies make designing and deploying these models challenging.Additionally,deploying TinyML models on resource-constrained hardware with specific software frameworks has broadened EdgeML’s applications across various sectors.These factors underscore the necessity for a comprehensive literature review,as current reviews do not systematically encompass the most recent findings on these topics.Consequently,it provides a comprehensive overview of state-of-the-art techniques in model conversion,inference mechanisms,learning strategies within EdgeML,and deploying these models on resource-constrained edge devices using TinyML.It identifies 90 research articles published between 2018 and 2025,categorizing them into two main areas:(1)model conversion,inference,and learning strategies in EdgeML and(2)deploying TinyML models on resource-constrained hardware using specific software frameworks.In the first category,the synthesis of selected research articles compares and critically reviews various model conversion techniques,inference mechanisms,and learning strategies.In the second category,the synthesis identifies and elaborates on major development boards,software frameworks,sensors,and algorithms used in various applications across six major sectors.As a result,this article provides valuable insights for researchers,practitioners,and developers.It assists them in choosing suitable model conversion techniques,inference mechanisms,learning strategies,hardware development boards,software frameworks,sensors,and algorithms tailored to their specific needs and applications across various sectors.
文摘Developing efficient neural network(NN)computing systems is crucial in the era of artificial intelligence(AI).Traditional von Neumann architectures have both the issues of"memory wall"and"power wall",limiting the data transfer between memory and processing units[1,2].Compute-in-memory(CIM)technologies,particularly analogue CIM with memristor crossbars,are promising because of their high energy efficiency,computational parallelism,and integration density for NN computations[3].In practical applications,analogue CIM excels in tasks like speech recognition and image classification,revealing its unique advantages.For instance,it efficiently processes vast amounts of audio data in speech recognition,achieving high accuracy with minimal power consumption.In image classification,the high parallelism of analogue CIM significantly speeds up feature extraction and reduces processing time.With the boosting development of AI applications,the demands for computational accuracy and task complexity are rising continually.However,analogue CIM systems are limited in handling complex regression tasks with needs of precise floating-point(FP)calculations.They are primarily suited for the classification tasks with low data precision and a limited dynamic range[4].
文摘Published proof test coverage(PTC)estimates for emergency shutdown valves(ESDVs)show only moderate agreement and are predominantly opinion-based.A Failure Modes,Effects,and Diagnostics Analysis(FMEDA)was undertaken using component failure rate data to predict PTC for a full stroke test and a partial stroke test.Given the subjective and uncertain aspects of the FMEDA approach,specifically the selection of component failure rates and the determination of the probability of detecting failure modes,a Fuzzy Inference System(FIS)was proposed to manage the data,addressing the inherent uncertainties.Fuzzy inference systems have been used previously for various FMEA type assessments,but this is the first time an FIS has been employed for use with FMEDA.ESDV PTC values were generated from both the standard FMEDA and the fuzzy-FMEDA approaches using data provided by FMEDA experts.This work demonstrates that fuzzy inference systems can address the subjectivity inherent in FMEDA data,enabling reliable estimates of ESDV proof test coverage for both full and partial stroke tests.This facilitates optimized maintenance planning while ensuring safety is not compromised.
基金supported by the National Natural Science Foundation of China(No.71501183).
文摘In order to solve the problems of high experimental cost of ammunition,lack of field test data,and the difficulty in applying the ammunition hit probability estimation method in classical statistics,this paper assumes that the projectile dispersion of ammunition is a two-dimensional joint normal distribution,and proposes a new Bayesian inference method of ammunition hit probability based on normal-inverse Wishart distribution.Firstly,the conjugate joint prior distribution of the projectile dispersion characteristic parameters is determined to be a normal inverse Wishart distribution,and the hyperparameters in the prior distribution are estimated by simulation experimental data and historical measured data.Secondly,the field test data is integrated with the Bayesian formula to obtain the joint posterior distribution of the projectile dispersion characteristic parameters,and then the hit probability of the ammunition is estimated.Finally,compared with the binomial distribution method,the method in this paper can consider the dispersion information of ammunition projectiles,and the hit probability information is more fully utilized.The hit probability results are closer to the field shooting test samples.This method has strong applicability and is conducive to obtaining more accurate hit probability estimation results.
基金supported by the National Natural Science Foundation of China(Nos.62176122 and 62061146002).
文摘Federated Graph Neural Networks (FedGNNs) have achieved significant success in representation learning for graph data, enabling collaborative training among multiple parties without sharing their raw graph data and solving the data isolation problem faced by centralized GNNs in data-sensitive scenarios. Despite the plethora of prior work on inference attacks against centralized GNNs, the vulnerability of FedGNNs to inference attacks has not yet been widely explored. It is still unclear whether the privacy leakage risks of centralized GNNs will also be introduced in FedGNNs. To bridge this gap, we present PIAFGNN, the first property inference attack (PIA) against FedGNNs. Compared with prior works on centralized GNNs, in PIAFGNN, the attacker can only obtain the global embedding gradient distributed by the central server. The attacker converts the task of stealing the target user’s local embeddings into a regression problem, using a regression model to generate the target graph node embeddings. By training shadow models and property classifiers, the attacker can infer the basic property information within the target graph that is of interest. Experiments on three benchmark graph datasets demonstrate that PIAFGNN achieves attack accuracy of over 70% in most cases, even approaching the attack accuracy of inference attacks against centralized GNNs in some instances, which is much higher than the attack accuracy of the random guessing method. Furthermore, we observe that common defense mechanisms cannot mitigate our attack without affecting the model’s performance on mainly classification tasks.
基金the Scientific Research Innovation Capability Support Project for Young Faculty(ZYGXQNJSKYCXNLZCXM-H8)Fundamental Research Funds for the Central Universities(2024ZYGXZR077)+3 种基金Guangdong Basic and Applied Basic Research Foundation(2023B1515120006)Guangzhou Basic and Applied Basic Research Foundation(2024A04J5776)the Research Fund(2023QN10Y421)Guangzhou Talent Recruitment Team Program(2024D03J0004),all related to this study.
文摘Osteoporosis is a known risk factor for rotator cuff tears(RCTs),but the causal correlation and underlying mechanisms remain unclear.This study aims to evaluate the impact of osteoporosis on RCT risk and investigate their genetic associations.Using data from the UK Biobank(n=457871),cross-sectional analyses demonstrated that osteoporosis was significantly associated with an increased risk of RCTs(adjusted OR[95%CI]=1.38[1.25–1.52]).A longitudinal analysis of a subset of patients(n=268117)over 11 years revealed that osteoporosis increased the risk of RCTs(adjusted HR[95%CI]=1.56[1.29–1.87]),which is notably varied between sexes in sex-stratified analysis.Causal inference methods,including propensity score matching,inverse probability weighting,causal random forest and survival random forest models further confirmed the causal effect,both from cross-sectional and longitudinal perspectives.
基金the financial support by the National Natural Science Foundation of China(Grant No.U24B2029)the Key Projects of the National Natural Science Foundation of China(Grant No.52334001)+1 种基金the Strategic Cooperation Technology Projects of CNPC and CUPB(Grand No.ZLZX2020-02)the China University of Petroleum,Beijing(Grand No.ZX20230042)。
文摘Offshore drilling costs are high,and the downhole environment is even more complex.Improving the rate of penetration(ROP)can effectively shorten offshore drilling cycles and improve economic benefits.It is difficult for the current ROP models to guarantee the prediction accuracy and the robustness of the models at the same time.To address the current issues,a new ROP prediction model was developed in this study,which considers ROP as a time series signal(ROP signal).The model is based on the time convolutional network(TCN)framework and integrates ensemble empirical modal decomposition(EEMD)and Bayesian network causal inference(BN),the model is named EEMD-BN-TCN.Within the proposed model,the EEMD decomposes the original ROP signal into multiple sets of sub-signals.The BN determines the causal relationship between the sub-signals and the key physical parameters(weight on bit and revolutions per minute)and carries out preliminary reconstruction of the sub-signals based on the causal relationship.The TCN predicts signals reconstructed by BN.When applying this model to an actual production well,the average absolute percentage error of the EEMD-BN-TCN prediction decreased from 18.4%with TCN to 9.2%.In addition,compared with other models,the EEMD-BN-TCN can improve the decomposition signal of ROP by regulating weight on bit and revolutions per minute,ultimately enhancing ROP.
基金supported by the National Key R&D Program of China(2022YFD1401600)the National Science Foundation for Distinguished Young Scholars of Zhejang Province,China(LR23C140001)supported by the Key Area Research and Development Program of Guangdong Province,China(2018B020205003 and 2020B0202090001).
文摘Inferring phylogenetic trees from molecular sequences is a cornerstone of evolutionary biology.Many standard phylogenetic methods(such as maximum-likelihood[ML])rely on explicit models of sequence evolution and thus often suffer from model misspecification or inadequacy.The on-rising deep learning(DL)techniques offer a powerful alternative.Deep learning employs multi-layered artificial neural networks to progressively transform input data into more abstract and complex representations.DL methods can autonomously uncover meaningful patterns from data,thereby bypassing potential biases introduced by predefined features(Franklin,2005;Murphy,2012).Recent efforts have aimed to apply deep neural networks(DNNs)to phylogenetics,with a growing number of applications in tree reconstruction(Suvorov et al.,2020;Zou et al.,2020;Nesterenko et al.,2022;Smith and Hahn,2023;Wang et al.,2023),substitution model selection(Abadi et al.,2020;Burgstaller-Muehlbacher et al.,2023),and diversification rate inference(Voznica et al.,2022;Lajaaiti et al.,2023;Lambert et al.,2023).In phylogenetic tree reconstruction,PhyDL(Zou et al.,2020)and Tree_learning(Suvorov et al.,2020)are two notable DNN-based programs designed to infer unrooted quartet trees directly from alignments of four amino acid(AA)and DNA sequences,respectively.
文摘Modern compression and acceleration methods for exploring efficient deep neural networks render real-world applications more feasible.Existing approaches uniformly apply the same procedure to every input image,overlooking instancewise complexity variations.Moreover,owing to pruning or decomposition techniques,the upper bound of network representation capabilities might be permanently diminished.In this work,an input-dependent multiscale dynamic inference method(MSDI)is developed to strike a better balance between model performance and inference acceleration.Specifically,we modify the main body of a convolutional network to obtain a series of parameter-sharing subnetworks with varying levels of complexity.A side branch structure is then introduced to assign an input instance to a suitable subnetwork as its inference route,and we expect to accelerate the inference by assigning the easy input to the subnetwork with low capacity.We further propose multiscale distillation training to optimize the training of the modified subnetworks.Additionally,we compare the entropy-based and learning-based grading approaches,aiming to obtain a more suitable route assignment method.Experiments show that MSDI can accelerate most existing convolutional models,achieving up to 74.7%computation savings across diverse datasets.
文摘Protocol Reverse Engineering(PRE)is of great practical importance in Internet security-related fields such as intrusion detection,vulnerability mining,and protocol fuzzing.For unknown binary protocols having fixed-length fields,and the accurate identification of field boundaries has a great impact on the subsequent analysis and final performance.Hence,this paper proposes a new protocol segmentation method based on Information-theoretic statistical analysis for binary protocols by formulating the field segmentation of unsupervised binary protocols as a probabilistic inference problem and modeling its uncertainty.Specifically,we design four related constructions between entropy changes and protocol field segmentation,introduce random variables,and construct joint probability distributions with traffic sample observations.Probabilistic inference is then performed to identify the possible protocol segmentation points.Extensive trials on nine common public and industrial control protocols show that the proposed method yields higher-quality protocol segmentation results.
文摘Chronic diseases are major causes of global death and disability,significantly impacting individual health and imposing economic burdens.This study aims to investigate the causal relationship between physical activity(PA)and the development of chronic diseases.Using a Mendelian randomization(MR)approach,we incorporated average PA and its subtypes(more than 450000 participants)as exposure measures and eight chronic diseases as outcome measures.Data were obtained from the European Genome-Wide Association Study(GWAS).The primary causal analysis technique employed was the inverse variance weighting(IVW)method,with MR-Egger,weighted median,weighted mode,and simple mode used to validate the results.Sensitivity analyses were performed to assess heterogeneity and pleiotropy.The IVW approach results show that vigorous physical activity(VPA)is associated with a modest reduction in the risk of major coronary heart disease(OR=0.95,95%CI:0.91–0.99,p=0.01).The causal directions of the other four MR methods are consistent with this result and validated by sensitivity analysis.No substantial associations were found between other levels of PA and chronic disease.Our findings underscore the importance of VPA in preventive cardiology and suggest its potential role in public health initiatives.Further research should explore the impact of PA on different demographic groups and the doseresponse relationship of VPA on heart health.
文摘Future 6G communications will open up opportunities for innovative applications,including Cyber-Physical Systems,edge computing,supporting Industry 5.0,and digital agriculture.While automation is creating efficiencies,it can also create new cyber threats,such as vulnerabilities in trust and malicious node injection.Denialof-Service(DoS)attacks can stop many forms of operations by overwhelming networks and systems with data noise.Current anomaly detection methods require extensive software changes and only detect static threats.Data collection is important for being accurate,but it is often a slow,tedious,and sometimes inefficient process.This paper proposes a new wavelet transformassisted Bayesian deep learning based probabilistic(WT-BDLP)approach tomitigate malicious data injection attacks in 6G edge networks.The proposed approach combines outlier detection based on a Bayesian learning conditional variational autoencoder(Bay-LCVariAE)and traffic pattern analysis based on continuous wavelet transform(CWT).The Bay-LCVariAE framework allows for probabilistic modelling of generative features to facilitate capturing how features of interest change over time,spatially,and for recognition of anomalies.Similarly,CWT allows emphasizing the multi-resolution spectral analysis and permits temporally relevant frequency pattern recognition.Experimental testing showed that the flexibility of the Bayesian probabilistic framework offers a vast improvement in anomaly detection accuracy over existing methods,with a maximum accuracy of 98.21%recognizing anomalies.
