arXiv subjects

Clara Grazian

Publications and source records attributed to Clara Grazian.

At least 19 recordsLinked to original sources

arXiv2026

Bayesian Copula Directional Dependence is Cross-Network Robust for Gene-Regulatory Pair Direction: A Benchmark Study on DREAM5

Inferring the direction of a gene-regulatory relationship is harder than inferring whether a relationship exists, and most direction-inference methods are validated mainly on a single in silico benchmark. We ask which method remains reliable as the data move from a synthetic network to real organisms and as sample size decreases. We embed a copula-based measure of directional dependence (CDD) in a Bayesian framework that returns, for each candidate pair, a posterior distribution over a directional contrast, a 95% credible interval, a posterior sign-support score, and a principled no-call. We benchmark this estimator against eight direction-inference methods, including two Bayesian DAG-posterior baselines, on the three core DREAM5 networks (in silico, S. aureus, and E. coli), with S. cerevisiae used as an out-of-domain eukaryotic stress test. Across the three core networks, Bayesian CDD is the only method whose called accuracy is always above 60%, whose coverage is always above 88%, and whose direction AUROC is always above 0.6; every competing method falls to chance or below on at least one network. CDD ranks first on both real-organism networks, remains stable on the smallest-sample network where bootstrap-interval methods collapse, and is the only Bayesian method that is simultaneously above chance and high-coverage under a 95% posterior gate. We position CDD as a post-screening, uncertainty-aware direction-refinement tool for candidate regulatory pairs.

stat.AP↗

arXiv2026

vPET-ABC: Fast Voxelwise Approximate Bayesian Inference for Kinetic Modeling in PET

Dynamic PET kinetic modeling increasingly demands voxelwise uncertainty quantification and robust model selection. Yet total-body PET (TB-PET) data volumes make conventional Bayesian approaches, such as per-voxel MCMC, computationally impractical, while deep models typically require retraining and careful revalidation when tracers, protocols, or kinetic models change, without necessarily improving inference speed. Vectorized voxelwise approximate Bayesian computation (vPET-ABC) is introduced as a likelihood-free, model-agnostic posterior inference framework for dynamic PET kinetic modeling at total-body scale. The method replaces explicit likelihood evaluation with forward simulations and a discrepancy test, then exploits full vectorization to transform voxelwise inference into an embarrassingly parallel workload suited to modern GPUs. In simulation, vPET-ABC produced posterior summaries with small divergence from sequential Monte Carlo baselines, and posterior mean estimates significantly more accurate than non-negative least squares (NNLS). For model selection between the linear parametric neurotransmitter model (lp-ntPET) and the multilinear reference tissue model, vPET-ABC maintained high sensitivity under high noise with moderate loss of specificity, whereas NNLS+Bayesian information criteria exhibited the opposite trade-off with near-zero sensitivity. In a human cigarette smoking dataset, vPET-ABC yielded denser probabilistic activation maps than lp-ntPET with effective number of parameters. On a 50 min total-body [18F]FDG study, vPET-ABC generated high quality whole volume K_i parametric images within practical runtimes on a single GPU, while also preserved local spatial correlation better than NNLS. Overall, vPET-ABC delivers fast, training-free, uncertainty-aware inference that scales to TB-PET and remains portable across tracers and kinetic models.

stat.AP↗

arXiv2025

Generative Flexible Latent Structure Regression (GFLSR) model

Latent structure methods, specifically linear continuous latent structure methods, are a type of fundamental statistical learning strategy. They are widely used for dimension reduction, regression and prediction, in the fields of chemometrics, economics, social science and etc. However, due to the lack of model inference, generative form, and unidentifiable parameters, most of these methods are always used as an algorithm, instead of a model. This paper proposed a Generative Flexible Latent Structure Regression (GFLSR) model structure to address this problem. Moreover, we show that most linear continuous latent variable methods can be represented under the proposed framework. The recursive structure allows potential model inference and residual analysis. Then, the traditional Partial Least Squares (PLS) is focused; we show that the PLS can be specialised in the proposed model structure, named Generative-PLS. With a model structure, we analyse the convergence of the parameters and the latent variables. Under additional distribution assumptions, we show that the proposed model structure can lead to model inference without solving the probabilistic model. Additionally, we proposed a novel bootstrap algorithm that enables uncertainty on parameters and on prediction for new datasets. A simulation study and a Real-world dataset are used to verify the proposed Generative-PLS model structure. Although the traditional PLS is a special case, this proposed GFLSRM structure leads to a potential inference structure for all the linear continuous latent variable methods.

stat.ME↗

arXiv2024

DARWIN 1.5: Large Language Models as Materials Science Adapted Learners

Materials discovery and design aim to find compositions and structures with desirable properties over highly complex and diverse physical spaces. Traditional solutions, such as high-throughput simulations or machine learning, often rely on complex descriptors, which hinder generalizability and transferability across different material systems. Moreover, These descriptors may inadequately represent macro-scale material properties, which are influenced by structural imperfections and compositional variations in real-world samples, thus limiting their practical applicability. To address these challenges, we propose DARWIN 1.5, the largest open-source large language model tailored for materials science. By leveraging natural language as input, DARWIN eliminates the need for task-specific descriptors and enables a flexible, unified approach to material property prediction and discovery. Our approach integrates 6M material domain papers and 21 experimental datasets from 49,256 materials across modalities while enabling cross-task knowledge transfer. The enhanced model achieves up to 59.1% improvement in prediction accuracy over the base LLaMA-7B architecture and outperforms SOTA machine learning approaches across 8 materials design tasks. These results establish LLMs as a promising foundation for developing versatile and scalable models in materials science.

cs.CL↗

arXiv2024

Approximate Bayesian Computation with Statistical Distances for Model Selection

Model selection in the presence of intractable likelihoods remains a central challenge in Bayesian inference. Approximate Bayesian computation (ABC) provides a flexible likelihood-free framework, but its use for model choice is known to be sensitive to the choice of summary statistics, often leading to poorly calibrated posterior model probabilities. Recent ABC variants based on statistical distances allow comparisons to be performed directly on empirical distributions, avoiding data reduction and offering improved theoretical guarantees under suitable conditions. This paper provides a systematic evaluation of discrepancy-based ABC methods for Bayesian model selection, focusing on their empirical behavior across a range of simulation settings and levels of model complexity. We compare full data ABC approaches based on Wasserstein, Creamer-von-Mises, and maximum mean discrepancy metrics with summary-statistic-based ABC and neural network classifiers. The results highlight settings in which full data ABC yields stable and well-calibrated posterior model probabilities, as well as scenarios where performance degrades due to model overlap or dependence. An application to toad movement models illustrates the practical implications of these findings. Overall, the study clarifies the strengths and limitations of discrepancy-based ABC for likelihood-free model choice and provides guidance for its use in realistic inferential settings.

stat.ME↗

arXiv2024

Novel Bayesian algorithms for ARFIMA long-memory processes: a comparison between MCMC and ABC approaches

This paper presents a comparative study of two Bayesian approaches - Markov Chain Monte Carlo (MCMC) and Approximate Bayesian Computation (ABC) - for estimating the parameters of autoregressive fractionally-integrated moving average (ARFIMA) models, which are widely used to capture long-memory in time series data. We propose a novel MCMC algorithm that filters the time series into distinct long-memory and ARMA components, and benchmarked it against standard approaches. Additionally, a new ABC method is proposed, using three different summary statistics used for posterior estimation. The methods are implemented and evaluated through an extensive simulation study, as well as applied to a real-world financial dataset, specifically the quarterly U.S. Gross National Product (GNP) series. The results demonstrate the effectiveness of the Bayesian methods in estimating long-memory and short-memory parameters, with the filtered MCMC showing superior performance in various metrics. This study enhances our understanding of Bayesian techniques in ARFIMA modeling, providing insights into their advantages and limitations when applied to complex time series data.

stat.ME↗

arXiv2024

Scalable Expectation Propagation for Mixed-Effects Regression

Mixed-effects regression models represent a useful subclass of regression models for grouped data; the introduction of random effects allows for the correlation between observations within each group to be conveniently captured when inferring the fixed effects. At a time where such regression models are being fit to increasingly large datasets with many groups, it is ideal if (a) the time it takes to make the inferences scales linearly with the number of groups and (b) the inference workload can be distributed across multiple computational nodes in a numerically stable way, if the dataset cannot be stored in one location. Current Bayesian inference approaches for mixed-effects regression models do not seem to account for both challenges simultaneously. To address this, we develop an expectation propagation (EP) framework in this setting that is both scalable and numerically stable when distributed for the case where there is only one grouping factor. The main technical innovations lie in the sparse reparameterisation of the EP algorithm, and a moment propagation (MP) based refinement for multivariate random effect factor approximations. Experiments are conducted to show that this EP framework achieves linear scaling, while having comparable accuracy to other scalable approximate Bayesian inference (ABI) approaches.

stat.ME↗

arXiv2024

State Space Modelling for detecting and characterising Gravitational Waves afterglows

We propose the usage of an innovative method for selecting transients and variables. These sources are detected at different wavelengths across the electromagnetic spectrum spanning from radio waves to gamma-rays. We focus on radio signals and use State Space Models, which are also referred to as Dynamic Linear Models. State Space Models (and more generally parametric autoregressive models) have been the mainstay of economic modelling for some years, but rarely they have been used in Astrophysics. The statistics currently used to identify radio variables and transients are not sophisticated enough to distinguish different types of variability. These methods simply report the overall modulation and significance of the variability, and the ordering of the data in time is insignificant. State Space Models are much more advanced and can encode not only the amount and significance of the variability but also properties, such as slope, rise or decline for a given time t. In this work, we evaluate the effectiveness of State Space Models for transient and variable detection including classification in time-series astronomy. We also propose a method for detecting a transient source hosted in a variable active galaxy, whereby the time-series of a static host galaxy and the dynamic nature of the transient in the galaxy are intertwined. Furthermore, we examine the hypothetical scenario where the target transient we want to detect is the gravitational wave source GW170817 (or similar).

astro-ph.GA↗

arXiv2024

SciQAG: A Framework for Auto-Generated Science Question Answering Dataset with Fine-grained Evaluation

We introduce SciQAG, a novel framework for automatically generating high-quality science question-answer pairs from a large corpus of scientific literature based on large language models (LLMs). SciQAG consists of a QA generator and a QA evaluator, which work together to extract diverse and research-level questions and answers from scientific papers. Utilizing this framework, we construct a large-scale, high-quality, open-ended science QA dataset containing 188,042 QA pairs extracted from 22,743 scientific papers across 24 scientific domains. We also introduce SciQAG-24D, a new benchmark task designed to evaluate the science question-answering ability of LLMs. Extensive experiments demonstrate that fine-tuning LLMs on the SciQAG dataset significantly improves their performance on both open-ended question answering and scientific tasks. To foster research and collaboration, we make the datasets, models, and evaluation codes publicly available, contributing to the advancement of science question answering and developing more interpretable and reasoning-capable AI systems.

cs.CL↗

arXiv2024

Bayesian Consistency for Long Memory Processes: A Semiparametric Perspective

In this work, we will investigate a Bayesian approach to estimating the parameters of long memory models. Long memory, characterized by the phenomenon of hyperbolic autocorrelation decay in time series, has garnered significant attention. This is because, in many situations, the assumption of short memory, such as the Markovianity assumption, can be deemed too restrictive. Applications for long memory models can be readily found in fields such as astronomy, finance, and environmental sciences. However, current parametric and semiparametric approaches to modeling long memory present challenges, particularly in the estimation process. In this study, we will introduce various methods applied to this problem from a Bayesian perspective, along with a novel semiparametric approach for deriving the posterior distribution of the long memory parameter. Additionally, we will establish the asymptotic properties of the model. An advantage of this approach is that it allows to implement state-of-the-art efficient algorithms for nonparametric Bayesian models.

stat.ME↗

arXiv2023

Stochastic Variational Inference for GARCH Models

Stochastic variational inference algorithms are derived for fitting various heteroskedastic time series models. We examine Gaussian, t, and skew-t response GARCH models and fit these using Gaussian variational approximating densities. We implement efficient stochastic gradient ascent procedures based on the use of control variates or the reparameterization trick and demonstrate that the proposed implementations provide a fast and accurate alternative to Markov chain Monte Carlo sampling. Additionally, we present sequential updating versions of our variational algorithms, which are suitable for efficient portfolio construction and dynamic asset allocation.

stat.CO↗

arXiv2023

DARWIN Series: Domain Specific Large Language Models for Natural Science

Emerging tools bring forth fresh approaches to work, and the field of natural science is no different. In natural science, traditional manual, serial, and labour-intensive work is being augmented by automated, parallel, and iterative processes driven by artificial intelligence-based experimental automation and more. To add new capabilities in natural science, enabling the acceleration and enrichment of automation of the discovery process, we present DARWIN, a series of tailored LLMs for natural science, mainly in physics, chemistry, and material science. This series relies on open-source LLM, incorporating structured and unstructured scientific knowledge from public datasets and literature. We fine-tuned the models using over 60,000 instruction data points, emphasizing factual correctness. During the fine-tuning, we introduce the Scientific Instruction Generation (SIG) model, automating instruction generation from scientific texts. This eliminates the need for manual extraction or domain-specific knowledge graphs and efficiently injects scientific knowledge into the model. We also explore multi-task training strategies, revealing interconnections between scientific tasks. DARWIN series not only achieves state-of-the-art results on various scientific tasks but also diminishes reliance on closed-source AI models. Our research showcases the ability of LLM in the scientific domain, with the overarching goal of fostering prosperity within the broader AI for science community.

cs.CL↗

arXiv2023

Clustering MIC data through Bayesian mixture models: an application to detect M. Tuberculosis resistance mutations

Antimicrobial resistance is becoming a major threat to public health throughout the world. Researchers are attempting to contrast it by developing both new antibiotics and patient-specific treatments. In the second case, whole-genome sequencing has had a huge impact in two ways: first, it is becoming cheaper and faster to perform whole-genome sequencing, and this makes it competitive with respect to standard phenotypic tests; second, it is possible to statistically associate the phenotypic patterns of resistance to specific mutations in the genome. Therefore, it is now possible to develop catalogues of genomic variants associated with resistance to specific antibiotics, in order to improve prediction of resistance and suggest treatments. It is essential to have robust methods for identifying mutations associated to resistance and continuously updating the available catalogues. This work proposes a general method to study minimal inhibitory concentration (MIC) distributions and to identify clusters of strains showing different levels of resistance to antimicrobials. Once the clusters are identified and strains allocated to each of them, it is possible to perform regression method to identify with high statistical power the mutations associated with resistance. The method is applied to a new 96-well microtiter plate used for testing M. Tuberculosis.

stat.AP↗

arXiv2023

Large Language Models as Master Key: Unlocking the Secrets of Materials Science with GPT

The amount of data has growing significance in exploring cutting-edge materials and a number of datasets have been generated either by hand or automated approaches. However, the materials science field struggles to effectively utilize the abundance of data, especially in applied disciplines where materials are evaluated based on device performance rather than their properties. This article presents a new natural language processing (NLP) task called structured information inference (SII) to address the complexities of information extraction at the device level in materials science. We accomplished this task by tuning GPT-3 on an existing perovskite solar cell FAIR (Findable, Accessible, Interoperable, Reusable) dataset with 91.8% F1-score and extended the dataset with data published since its release. The produced data is formatted and normalized, enabling its direct utilization as input in subsequent data analysis. This feature empowers materials scientists to develop models by selecting high-quality review articles within their domain. Additionally, we designed experiments to predict the electrical performance of solar cells and design materials or devices with targeted parameters using large language models (LLMs). Our results demonstrate comparable performance to traditional machine learning methods without feature selection, highlighting the potential of LLMs to acquire scientific knowledge and design new materials akin to materials scientists.

cs.CL↗

arXiv2023

Spatio-temporal stick-breaking process

Dirichlet processes and their extensions have reached a great popularity in Bayesian nonparametric statistics. They have also been introduced for spatial and spatio-temporal data, as a tool to analyze and predict surfaces. A popular approach to Dirichlet processes in a spatial setting relies on a stick-breaking representation of the process, where the dependence over space is described in the definition of the stick-breaking probabilities. Extensions to include temporal dependence are still limited, however it is important, in particular for those phenomena which may change rapidly over time and space, with many local changes. In this work, we propose a Dirichlet process where the stick-breaking probabilities are defined to incorporate both spatial and temporal dependence. We will show that this approach is not a simple extension of available methodologies and can outperform available approaches in terms of prediction accuracy. An advantage of the method is that it offers a natural way to test for separability of the two components in the definition of the stick-breaking probabilities.

math.ST↗

arXiv2023

Advances in Bayesian random partition models: A comprehensive review

Clustering is a crucial task in various domains of knowledge, including medicine, epidemiology, genomics, environmental science, economics, and visual sciences, among others. Methodologies for inferring the number of clusters have often been shown to be inconsistent, and incorporating a dependence structure among clusters introduces additional challenges in the estimation process. In a Bayesian framework, clustering is performed by treating the unknown partition as a random object and defining a prior distribution for it. This prior distribution can be induced by models assumed for the observations or directly defined on the partition itself. However, recent findings have revealed difficulties in consistently estimating the number of clusters and, consequently, the partition. Furthermore, summarizing the posterior distribution of the partition remains an open problem due to the high dimensionality of the partition space. This study aims to review Bayesian approaches for random partition models, highlighting the advantages and disadvantages of each method, and suggesting potential avenues for future research.

stat.ME↗

arXiv2023

An application of copulas to OPEC's changing influence on fossil fuel prices

This work examines how the dependence structures between energy futures asset prices differ in two periods identified before and after the 2008 global financial crisis. These two periods were characterised by a difference in the number of extraordinary meetings of OPEC countries organised to announce a change of oil production. In the period immediately following the global financial crisis, the decrease in oil prices and oil and gas demand forced OPEC countries to make frequent adjustments to the production of oil, while, since the first quarter of 2010, the recovery led to more regular meetings, with only three organised extraordinary meetings. We propose to use a copula model to study how the dependence structure among energy prices changed among the two periods. The use of copula models allows to introduce flexible and realistic models for the marginal time series; once marginal parameters are estimated, the estimates are used to fit several copula models for all asset combinations. Model selection techniques based on information criteria are implemented to choose the best models both for the univariate asset prices series and for the distribution of co-movements. The changes in the dependence structure of couple of assets are investigated through copula functionals and their uncertainty estimated through a bootstrapping method. We find the strength of dependence between asset combinations considerably differ between the two periods, showing a significant decrease for all the pairs of assets.

stat.AP↗

arXiv2023

Skew-Normal Posterior Approximations

Many approximate Bayesian inference methods assume a particular parametric form for approximating the posterior distribution. A multivariate Gaussian distribution provides a convenient density for such approaches; examples include the Laplace, penalized quasi-likelihood, Gaussian variational, and expectation propagation methods. Unfortunately, these all ignore the potential skewness of the posterior distribution. We propose a modification that accounts for skewness, where key statistics of the posterior distribution are matched instead to a multivariate skew-normal distribution. A combination of simulation studies and benchmarking were conducted to compare the performance of this skew-normal matching method (both as a standalone approximation and as a post-hoc skewness adjustment) with existing Gaussian and skewed approximations. We show empirically that for small and moderate dimensional cases, skew-normal matching can be much more accurate than these other approaches. For post-hoc skewness adjustments, this comes at very little cost in additional computational time.

stat.ME↗