Papers

We study optimal estimation of the location parameter of a distribution known only to lie in a symmetric moment class $\mathcal C_0$: the mean-zero distributions with bounded moment $\int\phi\, d\mathbb P\le B$ for a fixed even $\phi$. Our main result concerns the fixed-margin regime, where the error margin $\Delta$ is fixed as $n\to\infty$: we give an exact large-deviation characterization of the smallest worst-case probability $\beta_n(\Delta)$ of an error exceeding $\Delta$ that any measurable estimator can guarantee with $n$ observations. Its exponential rate is exactly a two-point Hellinger exponent over the class shifted to means $\pm\Delta$, $r(\Delta)=-\log\sup_{\mathbb P_{\pm\Delta}\in\mathcal C_{\pm\Delta}}\int\sqrt{d\mathbb P_{-\Delta}\, d\mathbb P_{\Delta}}$, achieved non-asymptotically, $\beta_n(\Delta)\le e^{-nr(\Delta)}$, by a monotone $M$-estimator synthesized from a two-parameter convex program. Lagrangian duality collapses the infinite-dimensional search over estimating functions to two multipliers, which determine a pair of envelopes characterizing the optimal estimating functions; the sandwich shape posited ad hoc in prior constructions emerges naturally. For bounded variance ($\phi(x)=x^2$, $B=\sigma^2$) the exponent is $r(\Delta)=\tfrac12\log(1+\Delta^2/\sigma^2)$. In the fixed-confidence regime, holding $\beta$ fixed and letting the optimal margin $\Delta_n(\beta)$ shrink with $n$, the same synthesis stays optimal to leading order for several concrete classes. As $\beta\downarrow0$ it attains the sharp constant $\sqrt2$ of Catoni for bounded variance and the constant $L(\alpha)$ of Lee and Bhatt et al. for bounded $\alpha$-moments, $\alpha\in(1,2)$, thereby shown tight; for slowly varying $\phi$ it is leading-order minimax at every fixed $\beta$. The least-favorable distributions are simple, supported on at most three atoms.

We use web search data to construct monthly indexes of derived demand for cobalt, copper, and nickel, which are key inputs in technologies driving the energy and digital transitions. We incorporate these indexes into Structural Vector Autoregressive (SVAR) models of global metal markets and identify structural shocks using zero, sign, and magnitude restrictions. This approach disentangles supply shocks from several demand-side drivers of metal prices and isolates a transition demand (TD) shock linked to the diffusion of metal-intensive technologies. We find that TD shocks generate persistent price effects, especially for copper and nickel, whereas supply and metal-specific demand shocks are more immediate and less persistent.

In modern applications of linear mixed models, the number of candidate fixed-effects covariates can grow exponentially with the sample size, while dependence induced by random effects and possible data contamination pose substantial challenges for existing variable screening methods. We propose a robust and computationally efficient sure screening procedure for identifying relevant fixed-effects covariates in ultrahigh-dimensional linear mixed models with known random effects. The proposed method leverages a proxy-based transformation to decouple dependence induced by random effects, enabling screening via marginal analysis in a transformed regression model. Robustness is achieved by constructing marginal utilities based on minimum density power divergence, yielding stability under data contamination and model misspecification without sacrificing scalability. The resulting procedure, termed DPD-SISP, is shown to retain all relevant covariates (sure screening property) with exponentially high probability under general conditions, allowing for non-Gaussian errors and nonpolynomial growth of dimensionality. In addition, DPD-SISP exhibits strong robustness properties supported by influence function and breakdown point analyses. The framework is further extended to incorporate prior information through conditional screening, mitigate correlation-induced masking via iterative refinement, and enable robust post-screening estimation of fixed effects. Extensive simulation studies demonstrate competitive performance of DPD-SISP under ideal settings and substantial gains in stability under data contamination. Its practical utility is illustrated through an application to high-dimensional longitudinal data from the ADNI2 study. The proposed framework thus provides a unified, robust, and scalable approach for variable screening in ultrahigh-dimensional linear mixed models.

We analyze the asymptotics of a block-Wishart random matrix ensemble of the type ${\boldsymbol W}_k = ({\boldsymbol X}^* \otimes {\boldsymbol I}_k){\boldsymbol T}({\boldsymbol X}\otimes{\boldsymbol I}_k)$ for ${\boldsymbol X} \in\mathbb{C}^{n\times p}$ with i.i.d. rows satisfying a suitable concentration-of-measure property, and ${\boldsymbol T} := \textrm{\bf Diag}({\boldsymbol T}_i)_{i\in[n]}$ a block diagonal matrix with self-adjoint blocks ${\boldsymbol T}_i\in \mathbb{C}^{k\times k}$, under the proportional asymptotics $n/p\to\alpha$ with $k$ fixed. These matrices play a prominent role in the analysis of $k$-index models in high-dimensional statistics. By studying the matrix Stieltjes transform of this random matrix model and its inverse ($K$-transform), we derive variational formulas for two functionals of the asymptotic spectral density of ${\boldsymbol W}_k$: the left (equivalently right) edge of its support, and its logarithmic potential.

We introduce a neural network-based framework for learning time series estimators through a process we term decision-theoretic pretraining. Analysts specify a generative world, a distribution over data-generating processes, and a target decision objective. A neural network trained on stratified simulations from this world approximates the corresponding optimal decision rule, yielding a neural estimator that provides forecasts, parameter estimates, predictive intervals, or model-selection for zero-shot inference on previously unseen time series. The joint specification of the generative world and objective enables the estimators to directly approximate process-level, finite-sample properties: near-optimal risk, bias control, minimax performance, and uniform calibration. Our experiments demonstrate that these neural estimators can outperform traditional baselines such as maximum likelihood estimation and model selection via AICc, for the same model structural model classes. Furthermore, even when trained purely on simulations of structural models, they achieve competitive or state-of-the-art forecasting accuracy on major real-world benchmarks, compared with statistical, neural or large pre-trained models. We illustrate the framework by addressing two longstanding challenges: finite-sample bias and miscalibration in AR(p) models, and the forecast combination puzzle. These applications highlight the approach's main advantage: its ability to approximate solutions to analytically intractable or computationally prohibitive time series problems, including complex structural equations or optimality criteria. Ultimately, by enabling explicit control over decision-theoretic trade-offs, the framework equips analysts with highly efficient estimation tools tailored to their specific analytical needs.

Pervasive data contamination – stemming from measurement errors, outliers, or adversarial corruption – has motivated the development of robust statistical methods. In this context, we propose a two-stage Adversarial Contamination-resistant Iterative Hard Thresholding (AC-IHT) algorithm for high-dimensional regression with contamination. Our nonconvex algorithm achieves minimax near-optimal (up to logarithmic terms) estimation by iteratively updating the coefficient vector and the contamination vector with different thresholding scales. We further demonstrate that our AC-IHT estimator is signal-adaptive: under proper signal conditions, it adaptively attains a sharper estimation rate and more accurate support recovery. Moreover, it enjoys the strong oracle property, laying a theoretical foundation for asymptotic inference. Numerical experiments confirm its superior finite-sample performance. Finally, we discuss theoretical extensions of the proposed procedure to generalized linear models and to heavy-tailed noise settings.

Switchback experiments and other clustered randomized designs are widely used on online platforms, but the clustered, time-dependent nature of these designs can make standard variance reduction methods behave differently than in standard A/B tests. We evaluate design-aware variance reduction methods for switchbacks – CUPED, CUPAC (ML-based covariate adjustment), and doubly robust (DR) estimators – relative to a baseline switchback analysis with cluster-robust standard errors. Through a hierarchical simulation framework that varies key regime parameters – number of clusters, cluster-size imbalance, within-cluster autocorrelation, carryover, and predictive signal strength – we evaluate validity (false positive rate and confidence interval coverage) and efficiency (standard error reduction, power, and minimum detectable effect as a function of run length). We also include a sensitivity analysis for cross-cluster spillovers to quantify bias and inference degradation under mild interference. The primary outcome is a practitioner-oriented regime map: when CUPED, CUPAC, or DR are most beneficial, and when time and cluster dependence and finite-cluster effects limit improvements.

High-dimensional educational datasets often exhibit sparsity, grouped predictors, and locally correlated covariates, limiting the effectiveness of conventional regression methods. We propose an Adaptive Weighted Group Fused LASSO estimator that jointly performs adaptive variable selection, group regularization, and coefficient fusion within a unified penalized regression framework. An efficient ADMM algorithm is developed, and asymptotic properties, including consistency, oracle property, and debiased asymptotic normality, are established. Simulation studies demonstrate superior estimation and prediction performance compared with existing penalized methods. An application to Alabama public school mathematics proficiency data illustrates improved model interpretability, predictive accuracy, and identification of the most influential institutional predictors.

Best subset selection procedures typically rely on a squared error loss, where a small number of outlying observations may distort the entire solution path. Replacing this loss with a robust alternative, such as the MM-estimator robust model selection criterion, is computationally prohibitive, as it requires an iterative fit across near-exhaustive candidate model spaces. To preserve the robust properties of the MM-estimator while avoiding its repeated iterative computation, we introduce FAMM, a Fast Approximate MM-Estimator based on a weighted least squares fit with weights obtained from a full data MM-estimator. Although the resulting estimator is no longer an MM-estimator, we prove that model selection consistency is retained when aligning the weights of the selection criterion with those of the estimator. We embed FAMM within COMBSS, whose continuous relaxation of the discrete best subset search enables scalable best subset selection for large numbers of predictor variables. Consequently, robust model selection and best subset selection with large and contaminated datasets are made feasible.

Structural changes often arise in real-world dynamic systems due to external interventions or environmental shifts, such as policy changes in epidemiology or climate forcing in environmental science. In this paper, we propose a unified framework for detecting and localizing structural changes in dynamic systems governed by ordinary differential equations. Unlike existing methods that assume mean or linear trend changes, our approach accommodates complex, nonlinear dynamics with both stable and diverging trajectories. We develop a new test statistic that combines residual-based discrepancy and normalized parameter contrast, capturing evidence for structural changes from both model fit and parameter shifts. Candidate structural changes are efficiently screened using a multiscale seeded-narrowest-over-threshold algorithm with a data-driven thresholding strategy. To refine selections and control false discoveries, we introduce a false discovery rate control procedure that leverages order-preserved sample splitting and symmetric contrast calibration. Theoretical guarantees are established, including detection consistency, near-minimax localization accuracy, and valid FDR control under weak dependence. Extensive simulations demonstrate superior performance over existing methods in both accuracy and FDR control. Applications to real-world data sets, including COVID-19 dynamics and global temperature trends, highlight the practical relevance and broad applicability of our method.

Forest aboveground biomass (AGB) is a critical indicator of ecosystem productivity and terrestrial carbon storage, yet regional carbon monitoring remains constrained by the sparse spatial and temporal availability of field inventories and airborne structural measurements. Recent Earth observation foundation models provide globally consistent geospatial representations derived from diverse multimodal datasets, offering a potential pathway toward scalable biomass monitoring. Here, we evaluate Google Satellite Embeddings (GSE), generated by the AplphaEarth Foundation Model, for regional-scale AGB estimation across diverse temperate forest ecosystems in the northeastern United States. We integrated annual GSE observations, airborne LiDAR, and continuous forest inventory measurements from the Northeastern Forest Inventory Network (NEFIN) within a machine-learning framework. Combined LiDAR-GSE models achieved an R^2 of 0.79 for AGB estimation. Capitalizing on annual GSE observations expanded the training dataset by more than tenfold through temporal growth adjustment, increasing predictive performance to R^2 = 0.82 while reducing model bias by over 70%. Spatial autocorrelation analyses showed that integrating foundation-model representations and structural predictors substantially reduced residual spatial dependence. Monte Carlo simulations demonstrated that hyperparameter optimization reduced model-performance variability by 27.9%. Our findings demonstrate that foundation-model Earth representations capture ecologically meaningful information relevant to forest biomass and provide a scalable framework for annual carbon monitoring in regions with incomplete airborne LiDAR coverage. Our fundings establish a pathway toward next-generation forest carbon assessment based on globally available foundation-model Earth observations.

Most approaches for accelerating Markov chain mixing either rely on incorporating expensive geometric information in the proposals, or reduce the per-step cost of sampling via surrogate densities. We propose a localisation principle that allows a surrogate-based Metropolis-Hastings proposal to exploit gradient-level geometric information of the target density, without evaluating either the target gradient or the surrogate gradient. The construction relies on regularisation and tempering of the proposal measure. We show that the expected proposal displacement coincides with the Langevin drift up to controlled error. The resulting framework, Delayed Acceptance with Regularisation and Tempering (DART), achieves an $O(\kappa \max\{\kappa, d\})$ mixing time from warm start for strongly log-concave targets with condition number $\kappa$ in $d$ dimensions. This matches the known $O(\kappa d)$ rate for MALA when $d \ge \kappa$, and scales as $O(\kappa^2)$, independent of dimension, otherwise. This is, to our knowledge, the first mixing time guarantee for a surrogate-transition-based MCMC method. We demonstrate DART on a hierarchical spatial generalised linear mixed model. In this setting, the Dirichlet-Neumann averaging parametrisation, originally introduced for the efficient simulation of Gaussian processes, is repurposed to supply the surrogate, and its linear memory and log-linear arithmetic scaling in the number of observation sites carry over to inference.

The 2024 Department of Justice antitrust complaint against RealPage, Inc. named five major residential REITs for coordinating algorithmic rent pricing across hundreds of thousands of apartment units in major US metropolitan areas. This paper studies whether census-tract-level corporate landlord concentration (CLC), measured from SEC EDGAR 10-K property filings geocoded to census tracts, the first such application in the literature, is associated with rent growth 2019-2023, and whether that association is larger in majority-minority neighborhoods. Rent outcomes are measured using the Zillow Observed Rent Index (ZORI). To account for the possibility that corporate landlords preferentially locate in neighborhoods already seeing rent appreciation, all regressions control for a fully novel Algorithmic Housing Burden Index (AHBI), a composite of pre-existing rent burden and market tightness from ACS data. Across 665 census tracts in ten US metropolitan areas, doubling REIT concentration is associated with 2.8 percentage points higher rent growth (p = 0.086, p = 0.030, HC1 robust). This association is significantly stronger in majority-minority tracts. Within the same metro, high-CLC majority-minority tracts are associated with 5.9 percentage points higher rent growth than comparable white tracts (p = 0.039). An XGBoost model predicts 44 percent of out-of-sample rent growth variance, with SHAP analysis independently confirming that CLC's contribution is positive in minority tracts and negative in white tracts. Taken all together, these findings provide the first tract-level evidence consistent with corporate landlord concentration being associated with disproportionately higher rent growth in communities of color.

Wearable devices can accurately measure human behavior, providing a unique opportunity to understand how behavior impacts health. Recent studies leveraging functional regression methods have found a strong relationship between accelerometer-collected physical activity and mortality. However, to determine if physical activity patterns impact mortality it is necessary to understand the causal effects of policies for physical activity, i.e., a function-valued treatment. Functional treatments present several challenges for causal effect estimation: 1) defining a scientifically meaningful estimand that reflects real-world policies and satisfies positivity is nontrivial; and 2) the potential for temporal confounding over continuous time. To address these, we propose stochastic policies for functional treatments that allow estimation of causal effects of changing the treatment distribution without requiring a positivity assumption. We develop a novel method for such that modifies the treatment through a single basis function chosen by the analyst, allowing for clear control over treatment modification and temporal confounding feedback. We show asymptotic normality of our estimators and that they exhibit rate double robustness. We apply our methods to the National Health and Nutrition Examination Survey to determine the causal effect of increasing physical activity over three-hour periods on mortality.

The global minimum-variance portfolio (GMVP) is the canonical decision built from an estimated covariance matrix, yet covariance estimators are universally evaluated by matrix-norm loss, which is not the object the decision depends on. We characterise exactly how covariance-estimation error maps into GMVP suboptimality. We prove an exact regret identity and a non-asymptotic bound showing decision regret depends on the estimation error only through its action on the portfolio weights, scaled by portfolio concentration and the conditioning of the true covariance. From this we derive the decision geometry: GMVP regret is invariant to a (p-1)-dimensional projection of the p^2-dimensional error matrix, with invariance to the covariance-scale direction as an exact special case. We then apply the framework to heavy-tailed returns (tail index kappa in (2,4)), establishing the regret convergence rate implied by the centred operator-norm rate, and confirm the theory on a skew-t/t-copula simulation design with pre-registered analysis. The decision-focused advantage is a sharper constant and a concentration discount rather than a faster rate; we report an honest high-conditioning boundary of the rate prediction. The results complement recent decision-focused learning approaches by supplying the exact estimation geometry and consistency theory they lack.

We address the problem of inferring a directed network from nodal measurements generated by linear diffusion dynamics on the sought graph. Observations are modeled as the outputs of a graph convolutional filter, i.e., a polynomial (with unknown coefficients) of a local diffusion graph-shift operator encoding the latent graph topology, excited with an ensemble of independent graph signals with arbitrarily-correlated nodal components. Unlike prior efforts that considered undirected graphs and white signal excitations, here the graph-shift operator and the observations' covariance matrix are not simultaneously diagonalizable. In this challenging context, we first rely on measurements of the output signals along with prior statistical information on the inputs to identify the diffusion filter. Such system identification problem involves solving a system of quadratic matrix equations, which we show is identifiable under spectral-diversity assumptions on the input covariances. For algorithmic purposes we recast it as a smooth quadratic minimization subject to Stiefel manifold constraints. Subsequent identification of the network topology given the graph filter estimate boils down to finding a sparse and structurally admissible shift that commutes with the given filter, thus, forcing the latter to be a polynomial in the sought graph-shift operator. A joint graph filter and topology identification algorithm is also proposed, which alternates between the aforementioned steps in a mutually reinforcing fashion to offer improved sample complexity. Numerical tests corroborate the effectiveness of the proposed algorithms in recovering synthetic digraphs and real-data case studies, and illustrate their potential utility on urban mobility analyses as well as portfolio optimization.

The morphology of a spectral-line profile contains information beyond scalar summaries of line strength, centroid, width, global asymmetry, or diagnostic line ratios. Broad wings, shoulders, double peaks, secondary components, and composite emission–absorption structures encode how flux is ordered across wavelength but can remain indistinguishable under conventional summaries. We introduce an interpretable geometric representation of line profiles inspired by rough path theory. Each wavelength-sampled profile is mapped to a common systemic rest-frame velocity grid and treated as a trajectory in velocity–flux space, traversed from blue to red. From this path, we define a compact set of low-order descriptors measuring signed velocity–flux area, blue–red imbalance localization, higher-order shape complexity, and emission–absorption ordering. Using synthetic profiles, we show that these descriptors separate morphologies with similar full width at half maximum (FWHM), non-parametric velocity width ($W_{80}$), and low-order moment summaries. We then apply the method to MaNGA integral-field spectroscopy by computing H$\alpha$ descriptors in individual spaxels and clustering them in a low-dimensional feature space. The resulting classes form spatially coherent regions of similar ordered line morphology. Although no external velocity field is supplied to the clustering, stacked spectra within these regions recover coherent large-scale centroid-velocity patterns broadly consistent with the MaNGA reference velocity fields. We release a minimalist MIT-licensed package ${\it spectropath}$, available at \href{https://rafaelsdesouza.com.br/spectropath/}{the project website}.

Many decoding methods for large language models can be understood as shifting probability mass toward outputs that are more likely under the model, either locally at the token level or globally at the sequence level. Therefore, their success depends on a fundamental question: when does sequence probability, that is, the conditional probability of a continuation given a prompt, actually align with correctness? In this paper, we set out to quantify this relationship across decoding methods, models, and benchmarks at four levels: across decoding methods, across hyperparameters within a method, across prompt-answer pairs within a dataset, and across repeated responses to the same prompt. We find that higher sequence probability is often predictive of correctness across prompt-answer pairs within a fixed dataset. However, this relationship does not generally transfer to decoding decisions: increasing sequence probability by changing hyperparameters or methods does not reliably improve accuracy. Further, sequence probability is not a good indicator of correctness for responses to the same prompt. These findings clarify when decoding can and cannot be expected to improve correctness, and provide practical guidance for decoding, self-consistency, and verifier-free self-improvement.

How different are several probability distributions from one another? For two distributions the standard answer is the family of Rényi divergences, singled out by two natural requirements: processing the data never makes distributions easier to tell apart, and independent repetitions add. Many problems in learning and statistics compare more than two distributions at once, such as testing among several hypotheses or bounding generalization against several priors. The same two requirements leave one kind of building block, built on a coincidence probability: how unlikely it is that independent samples, one from each distribution, all show the same empirical distribution. The logarithm is forced because repetitions add, which is already visible for a single experiment repeated. This characterization is known in greater generality, and this paper is about the meaning of its building blocks. On a finite alphabet, each building block indexed by a rational point of the simplex is the exponential rate of that coincidence as the samples grow in fixed proportions. Each is also the limiting free energy of Bayesian inference over distributions. At any amount of data, the free energy of the posterior is the coincidence measure plus two costs: the expected distance from a posterior draw to the most likely distribution, and the information gained per unit of data. Both costs vanish as data accumulate. When the comparison is conditioned on side information, every kind of building block has a conditional counterpart, and the coincidence ones alone do not suffice.

We study the fundamental problem of learning a high-dimensional Gaussian truncated to an unknown halfspace. Lee, Mehrotra and Zampetakis (FOCS'24) recently obtained the first polynomial time algorithm for this problem, but their resulting sample and time complexity bounds are not optimal. Under non-trivial truncation, for any target accuracy $\varepsilon > 0$ and dimension $d$ we give an efficient algorithm that uses $n = \tilde{O}(d^2/\varepsilon^2)$ samples and learns the underlying Gaussian to error $\varepsilon$ in total variation distance. Our algorithm is also fast: its runtime is dominated by the cost of computing the empirical covariance matrix. Both our sample and time complexity are optimal in terms of $d$ and $\varepsilon$ even without truncation: in this regard, we can learn a Gaussian under halfspace truncation for free. The key ingredient behind our result is a novel reinterpretation of the low-degree moments of the truncated Gaussian in terms of a relative truncation parameter. This relative truncation parameter uniquely determines the parameters of the untruncated Gaussian and enables direct parameter recovery. This reinterpretation allows us to circumvent the time intensive projected stochastic gradient descent procedure that is widely used in learning under truncation.

Reliably quantifying predictive uncertainty is difficult for complex, high-dimensional, or misspecified models. Both fully Bayesian and bootstrap resampling methods provide principled uncertainty estimates but are often too expensive for modern machine-learning models because they require posterior sampling or repeated model refitting. We introduce Ribbon, a scalable approximation to Dirichlet-reweighted bootstrap uncertainty. Ribbon replaces repeated refitting with an influence-function linearization around a single fitted model, preserving the first-order data-reweighting structure of the Bayesian bootstrap while requiring only post-hoc linear algebra. Ribbon approximates the Bayesian-bootstrap or weighted-likelihood-bootstrap refitting target. With a general concentration parameter, Ribbon gives a calibrated Dirichlet-reweighting family whose uncertainty scale can be tuned on validation data. We show that Ribbon is asymptotically equivalent to a flat-prior Laplace approximation under correct likelihood specification and recovers the robust sandwich covariance under misspecification. Across synthetic regression, MNIST classification, and California Housing benchmarks, Ribbon provides competitive predictive performance and improved calibration in several settings while avoiding repeated model retraining.

Training-free source selection for LLM families with shared vocabularies arises in scientific string domains such as SMILES, protein, and genomic sequences, where candidate corpora share a tokenizer but differ in prediction targets. This creates an activation-dark regime: representation-similarity metrics can be uninformative without assumptions about label-conditioned error geometry, while classical update-geometry metrics are computationally prohibitive at vocabulary scale. We show that, in a shared-output head setting, representation metrics (e.g., CKA) are non-identifiable for transfer; models can share identical representations yet have orthogonal head updates. The key identity is that head Fisher alignment is exactly a cosine between kernel mean embeddings in the joint activation-error space, exposing activation, error, and coupling factors rather than requiring a materialized Fisher matrix. FisherSketch estimates this cosine directly in a single streaming pass, making K=128,256 head Fisher alignment practical with a 16 KB task signature (m=4096) and a 192 KB per-task streaming state, small enough to store next to a model hash, but encoding transfer-relevant update structure. Beyond source selection, the same signatures and marginals provide a diagnostic instrument for studying whether LLM task similarity is driven by activations, errors, or their coupling; shared-parameter and internal-layer validations, together with Llama-3.1-8B verbalizer-shift experiments, show that FisherSketch remains informative when activation similarity cannot distinguish tasks.

This work addresses the problem of variance in stochastic gradient estimation for machine learning optimization. Deep learning relies on mini-batch methods such as stochastic gradient descent, which approximate full gradients but introduce noise, creating trade-offs between convergence stability, speed, and generalization. Existing methods, including variance reduction techniques (e.g., SVRG and SAG) and adaptive optimizers, aim to mitigate gradient noise but may introduce additional computational overhead. We propose a model-assisted sampling framework that interprets mini-batch gradients through survey sampling theory, treating the dataset as a fixed finite population and gradients as sample-based estimates. Our aim is to bridge machine learning optimization and survey sampling theory by combining their perspectives on sample-based estimation and variance reduction. By incorporating auxiliary gradient-prediction models, we construct more efficient gradient estimators, with uniform sampling arising as a special case when no auxiliary information is used. Our approach integrates easily with existing optimizers, improving efficiency without altering their dynamics. Empirical results on synthetic and six benchmark datasets show performance gains in 71-86% of the experiments, particularly for medium-sized input spaces in our benchmarks. Notably, with momentum-based optimizers such as AdamW, the proposed estimator achieves clearly better generalization in roughly half the training epochs compared to baseline estimator.

Graphical models for multivariate count data are widely used to characterize conditional dependence structures. For count variables with unbounded support, however, ensuring a finite normalizing constant typically imposes restrictive constraints on interaction parameters. We propose bounded discrete graphical models for multivariate discrete responses with finite support, which remove such constraints by construction while retaining interpretable dependence on the observed scale. We develop a regularized generalized score matching estimator (BRIDGE), which provides a normalization-free surrogate for likelihood-based estimation. The approach yields a unified system of estimating equations for all parameters and enables joint regularization through an $\ell_1$ penalty. To address degeneracy in the loss geometry, we introduce a reparameterization that restores curvature along the intercept direction and facilitates stable computation. On the theoretical side, we analyze a nonconvex objective and establish a population separation property that replaces global convexity. This yields nonasymptotic estimation error bounds and exact support recovery in high-dimensional regimes. Simulation studies and real data analyses demonstrate that BRIDGE accurately recovers graph structure and provides a stable and interpretable framework for high-dimensional discrete graphical modeling.

Global objectives, such as KL divergence and ELBO, are widely used in Bayesian inference for measuring distributional discrepancy. This paper studies distributional “local-mass behaviours” that are not directly captured by such global objectives. We introduce and use two mathematical tools: (1) Mass Index for recording the polynomial and logarithmic decay scales of local mass, and (2) regularised extended KL (RE-KL), a set-localised divergence that can be formulated in the presence of singular components. Mass Indices help characterise how Bayesian updating changes local mass: (1) power-log likelihood factors shift it explicitly, and (2) parameter-dependent supports, or their smooth softenings, may change the local scale through the amount of mass that remains near the parameter value. Using local RE-KL, we prove absolute, relative, and directional inequalities for comparing local small-ball masses under the two KL directions. Together, these results provide a local theoretical account of local mass behaviour. Experiments provide controlled illustrations of the local behaviour, and show that the directional comparison remains visible in the variational posteriors of Bayesian neural networks up to ResNet-50 on ImageNet. Code is available at https://github.com/Forsythia0604/Local-Mass-Framework.

Leibniz derivative estimation is a Monte Carlo technique for estimating derivatives of a discontinuous sample performance in stochastic models with respect to parameters of interest. By combining the push-out likelihood ratio (LR) method with Leibniz integral rules, it generalizes a broad class of existing LR-based derivative estimators. However, as an LR-based method, its variance is often higher than that of perturbation analysis-based methods and may grow linearly with the dimension of the stochastic input whose distribution depends on the parameter. In this paper, we propose a recursive conditioning approach and combine it with the Leibniz derivative estimation framework. The resulting conditional Leibniz estimator does not involve LR terms and therefore is not subject to variance growth with the input dimension. It also has a simple form and is easy to implement. We apply the method to an American call min-option model, and simulation results show its effectiveness and low-variance performance.

Uncertainty estimates in machine learning are typically evaluated using generic metrics such as the negative log-likelihood and expected calibration error, yet good performance on such metrics does not necessarily imply high utility in downstream decisions. We introduce decision-alignment, a criterion that reveals which evaluation metrics meaningfully align with downstream utilities. Applying this framework, we show that many widely used uncertainty metrics are either misaligned with common decision problems or encode pathological prior beliefs about the downstream task. We then propose prior-weighted utility metrics, a special class of proper scoring rules that provides decision-aligned uncertainty evaluation. Across benchmark experiments and real-world case studies, our metrics consistently align with realized decision utility, while conventional metrics do not. Our results surface flaws in the current UQ evaluation protocol and offer a principled extension of existing metrics toward decision-relevant UQ evaluation.

Climate extremes have become important drivers of mortality, producing sudden spikes that traditional mortality models fail to predict. To address this gap, we propose a two-step modelling framework that combines a regional weekly Lee-Carter baseline model that captures long-term mortality trends and overall seasonal patterns, with two complementary deep learning architectures designed to model excess mortality driven by environmental conditions and climate shocks. The first, a CNN-LSTM, captures region-specific temporal responses through convolutional filters. The second, a GNN-LSTM, replaces convolutions with graph-based representations to model spatial mortality dependencies and the propagation of climate-related impacts across regions. Both architectures are further extended to a quantile LSTM framework that produces time-varying prediction intervals. We evaluate our models against both the Lee-Carter baseline and MortFCNet (Zheng et al., 2025). Using French regional data over 1990-2019, our models capture delayed and nonlinear associations between environmental extremes and excess mortality. Both proposed architectures outperform the Lee-Carter baseline and MortFCNet across all regions, each reducing test MSE by approximately 24% relative to the MortFCNet, with particularly large gains at the oldest ages where climate-driven mortality spikes are most severe. From a risk management perspective, the proposed framework provides a more realistic characterization of extreme climate-driven mortality risk, with time-varying prediction intervals that offer a more informed basis for the assessment of climate-related longevity exposure by insurers and pension funds.

Empirical Bayes (EB) estimators can match the first-order asymptotic risk of maximum likelihood (ML) while behaving very differently at second order: recent excess mean squared error (XMSE) analysis shows that kernel-based EB estimation may be worse than ML when the kernel is poorly aligned with the true parameter. This paper turns that diagnostic into a design principle. We propose an XMSE-aware mixed estimator that interpolates between ML and EB shrinkage. Its fixed-weight XMSE is a scalar quadratic, yielding a closed-form oracle mixing weight that is no worse than both ML and the base EB estimator at the XMSE scale. A plug-in implementation based on finite-sample XMSE approximations is proved consistent, with a second-order oracle regret rate for an interior oracle weight. We further establish a transfer of the regret bound to the fixed-weight risk curve evaluated at the selected weight, a thresholded boundary rule, and extensions to compact kernel families and to finite and growing kernel dictionaries with high-probability oracle bounds. Finite impulse response simulations with SURE-tuned, hard-selection, and trace-corrected baselines, together with the public Silverbox and Cascaded Tanks benchmarks, show that the proposed estimator retains most of the benefit of regularization when it is helpful and retreats toward ML under kernel misspecification, with an identified finite-de analyzed on the benchmarks.

Whether in human, animal, or plant health, effective disease management requires the ability to characterize disease dynamics across space and time. In this context, integrating indirect indicators with broad spatio-temporal coverage, even when they are noisy, can provide valuable complementary information to direct measurements, which are often sparse because they are more costly or intrusive to collect. In this article, we propose a statistical framework to leverage such indirect indicators to predict disease severity at the individual or local-scale level and reconstruct large-scale disease dynamics. This two-step approach is able to account for the specific characteristics of disease severity observations, including zero inflation and spatio-temporal structure. The first step relies on a stacked hurdle model based on multiple random forests to locally predict disease severity from the available indirect indicators. In the second step a semi-parametric spatio-temporal model is used to reconstruct large-scale epidemiological dynamics over space and time from the indicators-based predictions. The proposed methodology is designed to be both generic and modular, and is illustrated by a case study in plant health. This case study focuses on the monitoring of sugar beet yellows disease in France between 2019 and 2023 by combining sparse field measurements and satellite-based remote sensing data.