Bayesian Inference Research Articles

A rapid multi-modal parameter estimation technique for LISA

Abstract The Laser Interferometer Space Antenna (LISA) will observe gravitational-wave signals from a wide range of sources, including massive black hole binaries. Although numerous techniques have been developed to perform Bayesian inference for LISA, they are often computationally expensive; analyses often take at least $\sim 1$ month on a single CPU, even when using accelerated techniques. Not only does this make it difficult to concurrently analyse more than one gravitational-wave signal, it also makes it challenging to rapidly produce parameter estimates for possible electromagnetic follow-up campaigns. simple-pe was recently developed to produce rapid parameter estimates for gravitational-wave signals observed with ground-based gravitational-wave detectors. In this work, we extend simple-pe to produce rapid parameter estimates for LISA sources, including the effects of higher order multipole moments. We show that simple-pe infers the source properties of massive black hole binaries in zero-noise at least $\sim 100\times$ faster than existing techniques; $\sim 12$ hours on a single CPU. We further demonstrate that simple-pe can be applied before existing Bayesian techniques to mitigate biases in multi-modal parameter estimation analyses of MBHBs.

Prognosis of reflective pavement cracking development with Bayesian updating and surrogate models

ABSTRACT The main objective of this study is to establish an integrated prognostic method for predicting the development of thermal-induced reflective cracking in asphalt overlay on concrete pavement. The integrated prognosis process for crack growth is developed on the modified Paris’ law and includes three steps: (1) calculation of J-integral; (2) Bayesian updating of crack growth parameters and (3) prediction of remaining life. Finite element modelling (FEM) was conducted to calculate J-integrals at different crack depths and cold weather conditions. A surrogate model was built based on the original Kriging method and it shows high accuracy in predicting J-integrals from maximum air temperature, temperature drop and crack depth. The Bayesian updating process starts from the laboratory-measured distribution of fracture model parameters (A and n) in Paris’ Law and converges to the true values to match the measured crack depths at different loading cycles from field inspection. On the other hand, the consideration of material damage in asphalt overlay avoids over-estimation of remaining loading cycles for reflective cracking to reach the pavement surface. Further study is recommended to accurately assess in-situ pavement modulus and detect crack depth over concrete joints for implementation of the proposed prognostics method.

Motion controller for multi-joint robotic arm with deep cascade gated Bayesian broad learning system

Intelligent controllers based on the broad learning system can simplify the process of model parameter adjustment, finding wide applications in the motion control of multi-joint robotic arms. However, motion controllers for multi-joint robotic arms based on broad learning system exhibit insufficient precision and overlook the impact of joint motion commonalities on controller design. Therefore, this paper proposes a novel motion control strategy for a multi-joint robotic arm based on a deep cascade feature-enhancement gated Bayesian broad learning system. Firstly, the motion controller of the deep cascade feature-enhancement Bayesian broad learning system is constructed to enhance the robotic arm motion control precision. Secondly, an incremental node generation module with an attention-gated mechanism is constructed to capture the unique motion characteristics of the target joints, which is further combined with model generalization to simplify the motion control process of the multi-joint robotic arm. Finally, controller convergence is enhanced by combining it with the Lyapunov theory to constrain the learning parameters. Simulations and physical experiments are designed to verify the feasibility and superiority of the proposed motion control strategy. The results demonstrated that the strategy improved the accuracy of robotic arm motion control, with the root mean square error in position tracking reduced to 0.0019 rad. This represents a 93.39% reduction in error compared to existing techniques.

Identifying causal relationships between 35 blood and urine biomarkers and urologic cancers: MR-meta combined with Bayesian colocalization Mendelian randomization analysis

BackgroundBlood and urine biomarkers have been associated with urologic tumors, but their causal relationship with urologic tumors is unclear.MethodsWe performed a bidirectional Mendelian randomization (MR) analysis of the association between 35 blood and urine biomarkers and urological tumors in our discovery cohort. A Bayesian weighting approach was used to validate the positive results identified in the discovery cohort, and steiger filtering analysis was used to distinguish causality from reverse causality. Bayesian colocalization analysis was used to analyze which single nucleotide polymorphisms (SNPs) specifically co-located between the positive blood and urine biomarkers and the disease phenotypes were driven, and MR-positive results from the discovery cohort and the validation cohort were combined using the MR-meta method.ResultsSeveral blood and urine biomarkers were found to be significantly and causally associated with urologic cancers. Notably, calcium (OR: 1.34, 95%CI 1.10–1.63, P = 0.0040) and sex hormone-binding globulin (SHBG) (OR: 0.81, 95%CI 0.70–0.95, P = 0.0092) were associated with bladder cancer; gamma glutamyl transferase (GGT) (OR: 0.91, 95%CI 0.83–0.99, P = 0.0209), lipoprotein A (Lp(a)) (OR: 1.12, 95%CI 1.01–1.24, P = 0.0399), and insulinlike growth factor 1 (IGF 1) (OR: 1.10, 95%CI 1.01–1.20, P = 0.0220) were linked to prostate cancer (PCa); non albumin protein (OR: 0.78, 95%CI 0.65–0.93, P = 0.0065) and total protein (OR: 0.80, 95%CI 0.64–0.99, P = 0.0380) were linked to renal cancer; and apolipoprotein A (Apo-A) (OR: 0.56, 95%CI 0.32–0.98, P = 0.0426) and urate (OR:1.89, 95%CI 1.03–3.47, P = 0.0399) were associated with renal pelvis cancer. These associations were validated in an independent cohort, with GGT, IGF 1, and Lp(a) being consistently linked to PCa.ConclusionThis study identified significant biomarkers associated with urological cancers in blood and urine. These include GGT, IGF 1, and Lp(a), which are strong biomarkers for PCa. In addition, the findings of this study provide evidence for a handful of risk and protective factors for the development of urologic cancers.

Traffic Flow Outlier Detection for Smart Mobility Using Gaussian Process Regression Assisted Stochastic Differential Equations

Current methods for detecting outliers in traffic streaming data often struggle to capture real-time dynamic changes in traffic conditions and differentiate between genuine changes and anomalies. This study proposes a novel approach to outlier detection in traffic streaming data that effectively addresses stochasticity and uncertainty in observations. The proposed method utilizes Stochastic Differential Equations (SDEs) and Gaussian Process Regression (GPR). By employing SDEs, we can capture drift and diffusion estimates in traffic streaming data, providing a more comprehensive modeling of the data generation process. Integrating GPR allows precise Bayesian posterior inferences for outlier detection within the SDE framework. To improve practicality, we introduce a flexible threshold-setting mechanism using statistical testing to control the false positive rate. This adaptability helps strike a balance between model fitting and complexity in outlier detection. Compared to traditional SDE-based methods, our SDE-GPR outlier detection method demonstrates enhanced robustness and better adaptability to the complexities of traffic systems. This is evidenced through an empirical study using time series data collected in California, USA. Overall, this study introduces a more advanced and accurate approach to outlier detection in traffic streaming data, paving the way for improved real-time traffic condition monitoring and management.

Variational inference for acceleration of SN Ia photometric distance estimation with BayeSN

Abstract Type Ia supernovae (SNe Ia) are standarizable candles whose observed light curves can be used to infer their distances, which can in turn be used in cosmological analyses. As the quantity of observed SNe Ia grows with current and upcoming surveys, increasingly scalable analyses are necessary to take full advantage of these new datasets for precise estimation of cosmological parameters. Bayesian inference methods enable fitting SN Ia light curves with robust uncertainty quantification, but traditional posterior sampling using Markov Chain Monte Carlo (MCMC) is computationally expensive. We present an implementation of variational inference (VI) to accelerate the fitting of SN Ia light curves using the BayeSN hierarchical Bayesian model for time-varying SN Ia spectral energy distributions (SEDs). We demonstrate and evaluate its performance on both simulated light curves and data from the Foundation Supernova Survey with two different forms of surrogate posterior – a multivariate normal and a custom multivariate zero-lower-truncated normal distribution – and compare them with the Laplace Approximation and full MCMC analysis. To validate of our variational approximation, we calculate the pareto-smoothed importance sampling (PSIS) diagnostic, and perform variational simulation-based calibration (VSBC). The VI approximation achieves similar results to MCMC but with an order-of-magnitude speedup for the inference of the photometric distance moduli. Overall, we show that VI is a promising method for scalable parameter inference that enables analysis of larger datasets for precision cosmology.

Unlocking Spanish pear genetic diversity: strategies for construction of a national core collection

Spanish pear germplasm collections are crucial for preservation, research, and breeding efforts. However, genetic diversity and structure is unknown at national level. A coordinated national project analyzed 1251 accessions from 7 Spanish pear collections using an internationally recognized set of 14 SSRs to enhance the utilization of these collections. Key findings included the identification of 760 unique genotypes (490 diploids and 270 triploids). Notably, genotypes represented by a single accession accounted for 49% of the total, indicating high vulnerability of this material. Using a Bayesian clustering method revealed two main genetic groups, G1 containing most foreign cultivars and G2 retaining local Spanish cultivars, which were further divided into two other subgroups using a nested approach, revealing moderate but significant differentiation among them. The populations were renamed according to the origin of the reference samples assigned to each group as ‘South’ (G1.1), ‘Western Europe-1’ (G1.2), ‘Western Europe-2’ (G2.1) and ‘No-Pyrus communis’ (G2.2). The results led to the creation of a ‘generalist’ collection, aiming to maximize genetic diversity representativeness, starting with 68 genotypes but expanding to 111 to achieve better allele recovery. This core collection is a valuable resource for genetic studies and conservation, enhancing efforts to preserve pear biodiversity.

Genetic variation and population structure of the rice accessions maintained in the AfricaRice genebank using DArTseq

AbstractUtilizing the full potential of rice collections mainly depends on an in‐depth exploration and understanding of the vast diversity in its germplasm. The AfricaRice genebank holds the largest collection of rice germplasm originating from the African continent. In the present study, we comprehensively characterized a collection of 9013 accessions, including Oryza barthii A. Chev., Oryza glaberrima Steud., Oryza longistaminata A. Chev. & Roehr., Oryza sativa L. ssp. indica, and Oryza sativa L. ssp. japonica, for genetic diversity and population structure using genotyping‐by‐sequencing through DArTseq analysis. We identified 27,718 high‐quality single nucleotide polymorphism markers after the genotypic data were filtered. Based on the analyses, the collection has extensive genetic diversity, and the average genetic distance of the entire set was 0.267 (range 0.001–0.469), with 45.1% of pairs of accessions being highly distant and 40.1% moderately distant from each other. Neighbor‐joining tree, principal component, and Bayesian population structure analyses clustered the 9013 accessions into six groups, based roughly on their taxonomic and biological status. The first, second, and third groups consisted of accessions belonging to O. glaberrima, O. barthii, and O. longistaminata, respectively. The fourth, fifth, and sixth groups were improved‐indica, japonica, and traditional‐indica accessions, respectively. The highest value of genetic variance proportion (PhiPT) was found in the species group followed by groups based on cluster analysis and on Bayesian population structure at K = 6. These results allow us to better understand the genetic diversity present in 9013 rice accessions maintained in the AfricaRice genebank and offer a valuable tool for pre breeding, breeding, and further genetic applications.

Mendelian Randomization Estimates the Effects of Plasma and Cerebrospinal Fluid Proteins on Intelligence, Fluid Intelligence Score, and Cognitive Performance.

Observational studies have revealed associations between levels of plasma and cerebrospinal fluid (CSF) proteins and cognition-related traits. However, these associations may be influenced by confounding factors inherent in observational research. This study aims to identify plasma and CSF proteins associated with intelligence, fluid intelligence score, and cognitive performance through the application of Mendelian randomization (MR). Proteomic quantitative trait locus (pQTL) data for plasma and CSF proteins were sourced from existing genome-wide association study (GWAS). Intelligence, fluid intelligence score, and cognitive performance GWAS summary statistics provided comprehensive data for two-sample MR analysis. Extensive sensitivity analyses, including Steiger testing, reverse MR analysis, and Bayesian co-localization, were conducted to validate associations and identify shared genetic variants. Phenotype scanning explored potential pleiotropic effects. MR analysis identified several proteins in plasma and CSF significantly associated with intelligence, fluid intelligence scores, and cognitive performance. For intelligence, negatively associated proteins in plasma include endoplasmic reticulum aminopeptidase 2 (ERAP2) and secretogranin III (SCG3), while positively associated proteins are myeloperoxidase (MPO), signal regulatory protein alpha (SIRPA), regulator of microtubule dynamics 1 (RMDN1), and endoplasmic reticulum lectin 1 (ERLEC1). In CSF, C1-esterase inhibitor and carboxypeptidase E (CBPE) both exhibited positive associations with intelligence. For fluid intelligence scores, negatively associated proteins in plasma are copine 1 (CPNE1) and SCG3, while positively associated proteins are nudix hydrolase 12 (NUDT12) and RMDN1.In CSF, Macrophage Stimulating Protein (MSP) demonstrated a significant negative impact. For cognitive performance, negatively associated proteins in plasma include ERAP2, tyrosine kinase with immunoglobulin-like and EGF-like domains 1 (TIE1), and SCG3, while positively associated proteins are NUDT12, RMDN1, ERLEC1, and ectonucleotide pyrophosphatase/phosphodiesterase family member 5 (ENPP5). In CSF, C1-esterase inhibitor was positively associated, while MSP and soluble tyrosine kinase with immunoglobulin-like and EGF-like domains 1(sTie-1) showed a negative association. Bayesian co-localization analysis revealed significant genetic overlaps between SIRPA, RMDN1, and ERLEC1 in plasma with intelligence; NUDT12 and SCG3 in plasma with fluid intelligence scores; and TIE1, NUDT12, RMDN1, ERLEC1, and ENPP5 in plasma with cognitive performance. Additionally, significant co-localization was identified between C1-esterase inhibitor and CBPE in CSF with intelligence, as well as between C1-esterase inhibitor and sTie-1 in CSF with cognitive performance. Reverse causality analysis confirmed the causal direction from proteins to cognitive traits. This study identifies specific plasma and CSF proteins that significantly impact intelligence, fluid intelligence scores, and cognitive performance. These proteins could serve as biomarkers and targets for future research and therapeutic interventions aimed at sustaining cognitive abilities and reducing impairment risks.

FLOW AND HEAT TRANSFER IN ROTATING COMPRESSOR CAVITIES WITH INVERTED SHROUD-THROUGHFLOW TEMPERATURE DIFFERENCES

Abstract In an aero-engine compressor, co-rotating discs form cavities that interact with an axial throughflow of secondary air at low radius. In the high-pressure (HP) compressor the shroud is hotter than the throughflow (directed downstream to the turbine) and the radial temperature gradient creates buoyancy-induced flow at Grashof numbers ∼ 1013. Such flows can be unstable and typically take the form of counter-rotating vortex pairs separated by radial hot and cold plumes. However, in low pressure (LP) and intermediate pressure (IP) compressors the secondary air is directed upstream. In this inverse scenario the axial throughflow is hotter than the compressor discs, reversing the disc temperature gradient and eliminating the fundamental driver for buoyancy. Despite its practical application and importance, this inverse scenario has not been previously investigated. The University of Bath Compressor Cavity Rig has been uniquely designed to simulate such flows, measuring temperature and unsteady pressure in the frame of reference of the rotating discs. Bayesian and spectral analysis have determined the radial distribution of disc heat flux, as well as the asymmetry of the rotating vortex structures and their slip relative to the discs. Unexpectedly, the new data reveal the flow structure in cavities with positive and inverted temperature differences are fundamentally similar (albeit with reversed radial-temperature profiles). Isothermal cases identified a critical Rossby number (Ro), above which the flow structure in the cavity was dominated by a toroidal vortex. At sub-critical Ro, the flow structure for the inverted temperature gradient continued

DNA sequencing reveals higher taxonomic diversity of coralline algae (Corallinales and Hapalidiales, Rhodophyta) in the tropical western North Atlantic that complicates ecological studies

Abstract Understanding of non-geniculate coralline algal (NGCA) diversity, prevalence, and distribution in the tropical western North Atlantic is currently limited by reliance on morphological and anatomical features that are inadequate for species identifications. Reef surveys from two study sites, the Florida Keys (Florida, USA) and Roatán (Honduras), provided prevalence data for the main benthic organisms as well as relative abundance for NGCA genera. NGCA collected during these surveys and from additional sites and depths were identified using DNA sequences from two plastid genes, psbA and rbcL, and two nuclear genes, SSU and LSU, and supported by morphological and anatomical descriptions. Bayesian and Maximum Likelihood phylogenetic analyses, along with species delimitation and genetic distance methods, resolved 11 genera within two orders. Within the Corallinales, the genera Dawsoniolithon, Porolithon and Spongites each had one species, Harveylithon and Lithophyllum had two, Titanoderma three, Hydrolithon four, Neogoniolithon eight, and an unknown genus within the subfamily Neogoniolithoideae had 11 putative species. Within the Hapalidiales, the genera Phymatolithon and Roseolithon each had one species. This study reinforces the inability of morphological and anatomical features to distinguish NGCA species in the tropical western North Atlantic, resulting in substantial underestimations of species richness and diversity.

Electrotactile proprioception training improves finger control accuracy and potential mechanism is proprioceptive recalibration

This study presents a novel training technique, visual + electrotactile proprioception training (visual + EP training), which provides additional proprioceptive information via tactile channel during motor training to enhance the training effectiveness. In this study, electrotactile proprioception delivers finger aperture distance information in real-time, by mapping frequency of electrical stimulation to finger aperture distance. To test the effect of visual + EP training, twenty-four healthy subjects participated in the experiment of matching finger aperture distance with distance displayed on screen. Subjects were divided to three groups: the first group received visual training and the other two groups received visual + EP training with or without a post-training test with electrotactile proprioception. Finger aperture control error was measured before and after the training (baseline, 15-min post, 24-h post). Experimental data suggest that both training methods decreased finger aperture control error at 15-min post-training. However, at 24-h post-training, the training effect was fully retained only for the subjects who received visual + EP training, while it washed out for the subjects with visual training. Distribution analyses based on Bayesian inference suggest that the most likely mechanism of this long-term retention is proprioceptive recalibration. Such applications of artificially administered sense have the potential to improve motor control accuracy in a variety of applications.

An efficient model updating method based on variational Bayesian inference with Wasserstein distance metric

An efficient model updating method based on variational Bayesian inference with Wasserstein distance metric

Inferring the distribution of the ionising photon escape fraction

The escape fraction of ionising photons from galaxies ($f_ esc $) is a key parameter for understanding how intergalactic hydrogen became reionised, but it remains mostly unconstrained. Measurements have been limited to the average value in galaxy ensembles and to handfuls of individual detections. To help understand which mechanisms govern ionising photon escape, here we infer the distribution of $f_ esc We developed a hierarchical Bayesian inference technique to estimate the population distribution of $f_ esc $ from the ratio of Lyman continuum to non-ionising UV flux measured from broadband photometry. We applied it to a sample of 148 $z 3.5$ star-forming galaxies from the VANDELS spectroscopic survey. We explored four physically motivated distributions: constant, log-normal, exponential, and bimodal, and recovered $ f_ esc for most models. We find the observations are best described by an exponential $f_ esc $ distribution with scale factor $ $. This indicates most galaxies in our sample exhibit very low escape fractions, while predicting substantial ionising photon leakage for only a few galaxies, implying a range of optical depths in the interstellar medium and/or time variability in ionising photon escape. We rule out a bimodal distribution at high significance, indicating that a purely bimodal model of ionising photon escape (due to very strong sightline and/or time variability) is not favoured. We compare our recovered exponential distribution with the SPHINX simulations and find that, while the simulation also predicts an exponential distribution, it significantly underpredicts our inferred mean. The distribution of $f_ esc $ can be a vital test for simulations in understanding ionising photon leakage, and is important to consider to gain a complete picture of reionisation.

Insight into the Phylogenetic Relationships of Phasmatodea and Selection Pressure Analysis of Phraortes liaoningensis Chen & He, 1991 (Phasmatodea: Lonchodidae) Using Mitogenomes

Stick and leaf insects are a group among the Insecta that are famous for their extraordinary mimicry ability. Since the establishment of the Phasmatodea, their internal classification has been constantly revised. Mitochondrial genes as molecular markers have been widely used for species classification, but the phylogenetic relationships within the Phasmatodea remain to be thoroughly discussed. In the present study, five mitogenomes of Phasmatodea ranging from 15,746 bp to 16,747 bp in length were sequenced. Bayesian inference (BI) and maximum likelihood (ML) analyses were carried out based on a 13 PCGs data matrix (nt123) and a combined matrix of 13 PCGs and two rRNA genes (nt123_rRNA). The present study supports the conclusion that Phylliidae was the basal group of Neophasmatodea and confirms the monophyly of Lonchodinae and Necrosciinae, but it shows that Lonchodidae was polyphyletic. A sister group of Bacillidae and Pseudophasmatidae was also recovered. The phylogenetic tree based on the nt_123 dataset showed higher node support values. The construction of a divergent time tree in this study supported the conclusion that extant Phasmatodea originated in the Jurassic (170 Mya) and most lineages diverged after the Cretaceous–Paleogene extinction event. To explore whether the mitochondrial genes of Phraortes liaoningensis collected from high latitudes where low temperatures occur for eight months of the year are under selection pressure, this study used the branch-site model and the branch model to analyze the selection pressure on the 13 mitochondria protein-coding genes (PCGs). We found that both ND2 and ND4L of Ph. liaoningensis exhibited positive selection sites using the branch-site model. This study shows that a low-temperature environment causes mitochondrial genes to be selected to meet the energy requirements for survival.

Advancing legal recommendation system with enhanced Bayesian network machine learning

Advancing legal recommendation system with enhanced Bayesian network machine learning

The complete mitochondrial genome of Dimorphopterus japonicus (Hidaka, 1959) (Hemiptera, Lygaeoidea) and phylogenetic relationships within the Lygaeoidea superfamily

Dimorphopterus japonicus is a kind of pest which seriously harms sorghum and millet. In this study, we sequenced the mitochondrial whole genome of D. japonicus (Hidaka, 1959), and characterized and analyzed the mitogenome. The D. japonicus genome length is 15,473 bp, and it exhibits a typical high A + T content (77.6%). The mitogenomic structure of D. japonicus is highly conservative and there are no gene rearrangements. The evolutionary rates of the PCGs are in the order of atp8 > nad3 > nad6 > nad2 > nad5 > atp6 > nad4 > nad4L > cox2 > nad1 > cytb > cox3 > cox1. By using the Bayesian inference and maximum likelihood methods, we inferred the phylogenetic relationships within Lygaeoidea and estimated their divergence times based on concatenated mitogenome genes. The stable clades of (Oxycarenidae + Piesmatidae), (Malcidae + Colobathristidae), (Meschiidae + Berytidae), and (Blissidae+(Cymidae + Ninidae)) were consistently recovered in all analyses. Estimated divergence times revealed that the divergence time of the Lygaeoidea was 99.4 Ma (95% HPD: 83.4-124.8 Ma), and most lygaeoid families diverged during the early Cretaceous to late Paleogene. Berytidae is the latest differentiation in the lygaeoid families, and the genus Dimorphopterus was differentiated in 45.9 Ma (95% HPD: 39.8–47.7 Ma). This study is of great significance for reconstructing the phylogeny of Lygaeoidea and providing insights into its evolutionary history.

Developing quantitative Adverse Outcome Pathways: An ordinary differential equation-based computational framework

The Adverse Outcome Pathway (AOP) biological framework was introduced in 2012, yet defining a mathematical/computational framework for quantitative AOP (qAOP) development remains an open problem. In order to properly unravel the intricate biological mechanisms described by AOPs and provide quantitative predictions to support risk assessment, a computational model should provide a clear time-course prediction of key events (KEs), as well as describe the key event relationships (KERs) linking a molecular initiating event (MIE) to an adverse outcome (AO). Ultimately, the mathematical description of those links entails the possibility of quantitatively predicting adverse effects based on early events.Here, we propose an ordinary differential equation (ODE) - based qAOP framework, as ODEs provide a time-course description of KEs and KERs. We illustrate how the application of computational techniques, such as Bayesian inference and Leave-one-out cross-validation (LOO-CV), can assist AOP development, introducing concepts of qAOP model selection and qAOP updating. Furthermore, we compare ODE and response–response based qAOP models, showing that ODE-based qAOPs can avoid erroneous predictions potentially resulting from response–response qAOPs. Finally, we show how ODE parameter variability can be linked to AO variability across a population. Overall, this framework serves as a valuable mathematical and computational tool for the development of qAOP models, enhancing our comprehension of intricate biological pathways associated with adverse outcomes.

Bayesian analysis of urban theft crime in 674 Chinese cities

Current academic research on fitting the volume of urban theft crimes at a macro size is limited, especially from the urban functionality perspective. Given this gap, this study utilizes a Bayesian model to conduct a fitting analysis of theft crime data from 674 cities in China from 2018 to 2020. This research aims to explore novel pathways for theft crime fitting. Results indicate that the size of urban functionality, particularly points of interest (POIs), exhibits excellent performance in fitting theft crimes, with POIs related to public services and commercial activities demonstrating the most significant fitting effects. This research successfully identifies effective indicators for crime fitting, thereby offering a new perspective and supplement to theft crime research. This study holds significant value for gaining a profound understanding of criminal phenomena and explaining the causes and mechanisms underlying the differences in theft crimes among various cities in China.

Bayesian deep learning: An enhanced AI framework for legal reasoning alignment

The integration of artificial intelligence into the field of law has penetrated the underlying logic of legal operations. Currently, legal AI systems face difficulties in representing legal knowledge, exhibit insufficient legal reasoning capabilities, have poor explainability, and are inefficient in handling causal inference and uncertainty. In legal practice, various legal reasoning methods (deductive reasoning, inductive reasoning, abductive reasoning, etc.) are often intertwined and used comprehensively. However, the reasoning modes employed by current legal AI systems are inadequate. Identifying AI models that are more suitable for legal reasoning is crucial for advancing the development of legal AI systems.Distinguished from the current high-profile large language models, we believe that Bayesian reasoning is highly compatible with legal reasoning, as it can perferm abductive reasoning, excel at causal inference, and admits the "defeasibility" of reasoning conclusions, which is consistent with the cognitive development pattern of legal professionals from apriori to posteriori. AI models based on Bayesian methods can also become the main technological support for legal AI systems. Bayesian neural networks have advantages in uncertainty modeling, avoiding overfitting, and explainability. Legal AI systems based on Bayesian deep learning frameworks can combine the advantages of deep learning and probabilistic graphical models, facilitating the exchange and supplementation of information between perception tasks and reasoning tasks. In this paper, we take perpetrator prediction systems and legal judegment prediction systems as examples to discuss the construction and basic operation modes of the Bayesian deep learning framework. Bayesian deep learning can enhance reasoning ability, improve the explainability of models, and make the reasoning process more transparent and visualizable. Furthermore, Bayesian deep learning framework is well-suited for human-machine collaborative tasks, enabling the complementary strengths of humans and machines.