Papers

We study poverty minimization via direct transfers, framing this as a statistical learning problem while retaining the information constraints faced by real-world programs. Using nationally representative household consumption surveys from 34 countries that together account for 76% of the world's poor, we estimate that reducing the poverty rate to 1% (from a baseline of 13%) would cost $211 B nominal per year. This is 4.0 times the corresponding reduction in the aggregate poverty gap, but only 19% of the cost of universal basic income. Extrapolated globally, the results imply a cost of 0.28% of global GDP to (approximately) end extreme poverty.

We study a variant of the Thompson Sampling (TS) algorithm, called $\alpha$-TS, for solving stochastic generalized linear bandit problems. Existing analyses of TS require inflating the posterior variance to derive near-optimal regret guarantees. We formalize the idea of variance inflation by introducing $\alpha$-TS that uses a fractional or $\alpha$-posterior instead of the standard posterior. Our main contribution is to identify general regularity conditions on the prior and reward distributions that enable a regret analysis of $\alpha$-TS without assuming any tractable approximation of the posterior distribution, unlike previous works. For a specific choice of $\alpha \propto d^{-1}$, our general regret bound yields the best known regret bound of $O(d^{3/2}\sqrt{T}\log T)$ for both the exponential and sub-Gaussian families of reward distributions. We further provide an $\alpha$-dependent lower bound showing that the regret constant depends on the product $\alpha d$, and that when $\alpha \propto d^{-1}$ the regret scales as $\Omega(d^{3/2}\sqrt{T})$, explaining the origin of the $d^{3/2}$ factor in the upper bound. Our proof technique adapts and combines recent advancements in the analysis of linear bandit problems with first- and second-order posterior concentration theory from the Bayesian statistics literature.

Time-varying treatment effects can be estimated with linear regression paired with jackknife confidence intervals. In experiments at a large online video platform, the test based on this estimator is not much less powerful than modern tests for monotonic trends. If you suspect trends in your online experiments, consider using this estimator.

Rare clinical outcomes pose a difficulty deeper than ordinary class imbalance: a penalized logistic model can return finite, stable-looking coefficients before the data support a reliable threshold decision. We formulate the accrual question as decision-targeted sequential certification. On a prespecified finite monitoring schedule and target set, asymptotic prediction bands are adjusted for simultaneous coverage, and certification is assessed only among profiles that might be referred, so that a large low-risk majority cannot trigger an uninformative stop. Under the working rare-event logistic model and stated regularity conditions, each certified decision is asymptotically model-conditionally correct with probability at least \(1-\alpha\) over the schedule, and the effective information scales with the number of genuine events. Simulations and a US linked birth/infant-death application illustrate the gap between a model that is merely estimable and one whose decisions are certifiable: the whole-population rule stopped while more than half of referral-relevant profiles remained ambiguous, whereas the decision-targeted rule did not certify by the 300{,}000-birth horizon despite stable temporal validation. Regularization makes a rare-event model estimable but does not substitute for genuine rare-event information.

To deal with superfluous clusters and to reduce the number of clusters as often desired in practice, a modal cluster merging procedure is proposed. Based on some new properties established in this paper for Morse functions, the procedure merges clusters in a sequential manner without causing unnecessary density distortion. Each cluster is evaluated for its significance relative to the other clusters, using the Kullback-Leibler divergence or its log-likelihood approximation, by truncating the density for the cluster at an appropriate level. The least significant cluster is then merged into one of its adjacent clusters, using the novel concept of cluster adjacency defined in this paper. The resulting hierarchical clustering tree is useful for determining the number of clusters, as may be preferred by a specific user or in a general, meaningful manner. Numerical studies show that the new procedure deals well with difficult clustering problems and often produces intuitively appealing and numerically more accurate clustering results, as compared with several other popular clustering methods in the literature.

Contemporary oncology drug development increasingly requires efficient dose-optimization strategies that evaluate monotherapy (Mono) and combination therapy (Combo) while balancing activity, efficacy, and tolerability. We propose a staggered seamless phase I/II design for settings in which a novel agent is evaluated alone and in combination with an established therapy. In phase I, Mono dose finding begins first, and Combo subtrials can be opened adaptively once a prespecified combination-initiation signal based on early clinical or biological information is observed. Dose assignment uses a model-assisted rule based on toxicity and early activity, with backfilling at tolerable and potentially promising regimens. At the end of phase I, two candidate regimens are selected from the evaluated Mono and Combo regimens using an efficacy-toxicity utility based on accumulated toxicity and treatment-response data. Phase II seamlessly carries forward patients treated at the selected regimens, enrolls additional patients as needed, and applies Bayesian futility and efficacy stopping boundaries to identify a final recommended optimal biological dose (OBD). Simulation studies showed that the proposed design shortened phase I trial duration relative to the comparator designs while maintaining competitive OBD-selection performance and acceptable safety. The seamless phase II component further reduced the need for additional enrollment and supported efficient final OBD selection.

Conditional quantiles are central to asymmetric decision losses, tail-risk assessment, and interval forecasts, but Bayesian quantile regression for nonlinear time series can be difficult when the temporal feature vector is high-dimensional. We develop the quantile deep echo state network (Q-DESN), a Bayesian quantile regression model conditional on fixed features generated by a deep echo state network. Conditional on a specified reservoir construction and feature map, posterior uncertainty is assigned to the regression coefficients, likelihood, and shrinkage parameters. Single-level fits use asymmetric Laplace or quantile-fixed generalized asymmetric Laplace working likelihoods with ridge or regularized-horseshoe priors. Posterior computation uses Markov chain Monte Carlo when computationally practical and a model-specific variational Bayes approximation for analyses requiring repeated fitting. For quantile grids, we compare independent level-wise regressions followed by monotone rearrangement with a joint quantile-vector regression that shrinks adjacent quantile-specific coefficient differences. Synthetic studies, a single-origin retrospective Global Flood Awareness System (GloFAS) streamflow case, and a retrospective PriceFM comparison identify settings where this fixed-feature Bayesian regression improves finite-grid quantile scores.

Here we present a method for measuring, analyzing and reducing theoretical uncertainty. We call it metatheoretical multiverse analysis (MMA). The method is important because the reliability and replicability of scientific observation and testing of a phenomenon are a function of theoretical uncertainty thus reducing it is a method for improving theory. The MMA method requires encoding theory as propositional logic between variables (nodes) and their relationships with one another (edges). Propositional logic models are treated as causal path models in our method, which enable mathematical properties of testing and identification underlying their causal propositions. By encoding theories-as-data including their unknown components, it is possible to generate a multiverse of alternatively plausible theoretical models which can then be meta-analyzed using techniques borrowed from (statistical) multiverse analysis. We introduce metrics for studying theoretical multiverses that enable measuring and identifying the causes of metatheoretical uncertainty. This guides researchers to where it is best to invest theoretical development to reduce uncertainty. We use three simulations and a dedicated software package to demonstrate this method.

Large spatial data sets now record many correlated variables at many thousands of locations, often on domains where Euclidean distance misrepresents proximity. The central difficulty is modelling the cross-variable dependence jointly while retaining variable-level interpretation. We introduce the bigraphical Mat'ern-Whittle process, a multivariate Gaussian process that resolves this with two graphs. A spatial graph generates the Mat'ern structure of each variable through a fractional power of a graph Laplacian, so the process is valid on any topology, with per-variable range, smoothness and amplitude. A directed acyclic variable graph encodes the scientific structure: we prove that each absent edge yields an exact conditional independence between the corresponding fields. We further prove that the operator determinant does not involve the cross-dependence coefficients, which keeps matrix-free likelihood evaluation and Bayesian learning of the variable graph tractable at scale. Estimation requires only sparse matrix-vector products and scales to tens of millions of space-variable pairs. In simulations the method recovered parameters and graphs accurately, remained robust under misspecification, and halved held-out prediction error on a non-convex domain. In a spatial transcriptomics section with 19,809 cells and 1,122 genes, fitted in 75 minutes on a laptop, borrowing across the learned gene graph reduced held-out prediction error by 50 to 91 percent. Theoretical challenges, such as the achievable efficiency of estimating the variance of the nugget, are also explored.

In incomplete markets, no-arbitrage (NFLVR) guarantees the existence, not the uniqueness, of an equivalent local martingale measure (ELMM): unhedgeable risks (jumps, stochastic volatility) admit a whole family of equivalent measures, and asset dynamics alone cannot pin down the one the market selects. We characterize the identifiability of $Q$ from option data and propose a measure that is identifiable from data yet prices any claim consistently. The key boundary is an “identification wall”: European options identify only the terminal marginal, while out-of-sample tails and path/joint structure require instruments matched to the priced risk (variance or higher-moment swaps, path-dependent claims). Within this view, minimum-relative-entropy weighted Monte Carlo (WMC) is the optimal baseline for the marginal; we generalize it to a full path-space measure change parameterized by a neural network on physical scenarios—XiNet—which learns $\xi=\mathrm{d}\mathbb{Q}/\mathrm{d}\mathbb{P}$ directly from 8 model-free path features, with option prices as a soft constraint. On European (marginal) pricing XiNet matches but does not surpass WMC, and both are bound by the identification wall out-of-sample. On path-dependent claims the picture reverses: calibrated on the same European surface, per-marginal methods fail structurally (an ATM forward-start is mispriced by $\sim+100\%$), whereas XiNet's single self-consistent measure keeps the bias to $+0.3\%$, beating maximum-entropy WMC ($-24\%$), because $\xi=f_\theta(\text{path features})$ captures joint structure Europeans cannot constrain. Identification is thus risk-specific, and XiNet is a single measure that absorbs available instruments and prices all claims consistently.

Understanding the generative mechanism of real-world networks is crucial for analyzing connection patterns and making inference from network data. Graphons are widely used to model such mechanisms. Network histogram methods are nonparametric approaches based on blockmodel approximations that provide an intuitive view of network connection structures. However, there is a lack of software packages that construct network histograms. We introduce the R package nethist, which implements network histogram-type graphon estimators within a unified interface. This package is applicable to both single-layer and multilayer networks, and includes graphical summaries for examining both local and global network structures. By providing comprehensive network analysis tools, nethist facilitates understanding of complex systems of interrelated vertices.

In this article, we address the problem of uncertainty quantification of state variables in the context of inverse problems. Inverse problems are associated with phenomena that can be represented through ordinary or partial differential equations, for which observations or data are available, but the values of the parameters that characterize the equations, the initial or boundary conditions are not known. Currently, the literature offers very few alternatives for analyze the propagation of uncertainty of state variables, which are limited to constructing pseudo-credible regions or quantifying their uncertainty at isolated points. We propose a methodology that combines tools of Bayesian inference, with Hamiltonian Monte Carlo sampling employed for efficient posterior exploration, and functional data analysis, specifically the Modified Band Depth method, to determine credible regions for the trajectories of the state variables throughout the time horizon of interest. We expose a methodology validation through a simulation study, which shows that our proposal captures a higher proportion of state variable trajectories than traditional pointwise analysis methods, our proposal generated credible regions that contained the true trajectory of the state variables 96.4% of the times, versus 80% that the regions of the pointwise method did. Furthermore, we demonstrate its application to a non-trivial model associated with a neurological phenomenon, for which the methodology effectively captures the time-dependent dynamics of the state variables.

Functional linear models are commonly used for analyzing functional data with scalar responses, particularly for modeling the conditional mean of the response given functional covariates. In this work, we extend the traditional scalar-on-function regression in two major aspects: 1. estimating the conditional distribution function instead of a particular characteristic such as the conditional mean; 2. considering an arbitrary operator without imposing functional linearity or any model assumption. We use a likelihood approach for the conditional hazard function and apply convolutional neural networks to approximate the effects of functional inputs and estimate the conditional distribution using deep operator networks. Through simulations and a real world data example, we show the desirable robustness of the proposed method in comparison with the mean regression neural networks, demonstrating that our approach achieves better conditional distribution estimation and interval prediction in complex data settings.

Many statistical models involve parameter-dependent normalizing constants that are computationally intractable, creating substantial obstacles to standard Bayesian inference. Although existing likelihood-based algorithms can often circumvent these constants, their uncertainty quantification may be poorly calibrated under model misspecification. To address these challenges, we propose SME-BETEL, a semiparametric Bayesian framework that combines score matching estimating equations with Bayesian exponentially tilted empirical likelihood. The resulting posterior avoids evaluation of normalizing constants and does not require learning-rate calibration. Building on this framework, we develop a new score matching criterion for mixed-domain data, extending SME-BETEL to models whose observations combine components from different sample spaces. This construction enables robust Bayesian inference for mixed-domain doubly-intractable models. We establish consistency and asymptotic normality of the score matching estimator, and prove a Bernstein-von Mises theorem for the SME-BETEL posterior. These results show that SME-BETEL credible sets are asymptotically calibrated to the sampling variability of the score matching estimator, yielding valid frequentist coverage under model misspecification. Simulation studies show that SME-BETEL remains competitive under correct specification and substantially improves uncertainty quantification under misspecification. An ozone-monitoring application demonstrates the practical utility of the mixed-domain construction for spatial preferential modeling.

A system often has to act long before it learns whether the act worked: a recommender sees a click in seconds and a purchase in days. With $K$ actions and a delay of $d$ rounds, the best rate known for this setting is $\widetilde{O}(\sqrt{(K+d)T})$ over $T$ rounds, so a longer menu is always more expensive to learn from. It need not be: if the outcome depends on the action only through the state it produced, then one late outcome informs every action that could have produced the observed state, and the price is set by how many genuinely different states the actions produce rather than by how many actions there are. We measure this using an effective dimension $v_t$ between $1$ and the number of states, and prove $\widetilde{O}(\sqrt{(d+1)V\log K})$ for a rotating algorithm and $\widetilde{O}(\sqrt{V^{-}}+\sqrt{dT})$ for the single-copy algorithm used in practice, for any budget fixed in advance; merging similar states lowers the price further, at an explicit bias. Even when given the exact losses from $d$ rounds ago, no algorithm escapes $\Omega(\sqrt{dE\min\{1+\log J,T/d\}})$, where $J$ counts the drifting directions and $E$ bounds how far losses move while the learner waits. On generated data, the state channel cuts regret by up to 79 percent against action-level weighting and, on the funnel family, by 32 to 68 percent against a tuned minimax-optimal method.

In many inverse problems, the aim is to recover a probability distribution on a latent (or object state) space from indirect, noisy observations. When the observations can be modelled as noisy samples from the mixing latent distribution, the recovery problem is a deconvolution problem on the space of probability measures. A common strategy is to fix a finite set of candidate points and estimate a weight for each, turning the problem into a finite-dimensional concave maximum likelihood problem on the simplex. We study the combined effect of this discretization and of the noise in the context of cryo-electron microscopy (cryo-EM), where the candidate points are biomolecular conformations and the weights describe the relative frequency of each conformation. Our results pertain to both statistical and algorithmic aspects of the estimator. We analyze the weight-recovery problem in which the candidate states and their likelihoods are known. A nearby pair of candidates forces a near-null direction. More generally, the grid and the noise level impose a uniform lower bound on the achievable Kullback–Leibler divergence between observation densities, even with infinite data. The finite-grid estimator is asymptotically normal when its population target is in the interior of the simplex; at a boundary target, its limit is a cone-projected Gaussian. Finally, the exact proximal form of Expectation–Maximization leads to a global high-noise comparison between an early iterate and a KL-penalized likelihood, without a basin assumption or a linearization of the recursion. Tests on synthetic images of the Hsp90 molecule illustrate the theoretical findings and translate them into practical guidelines for interpreting reweighted ensembles.

Accurate large-scale wildfire spread modelling requires models that capture both the environmental conditions associated with fire occurrence and the local dynamics of fire propagation. We propose a three-stage framework that combines a Random Forest (RF) model with a cellular automaton (CA). First, an RF model trained on the 2021 Canadian fire season estimates daily pixel-level fire-occurrence probabilities. Second, quantile gradient boosting models provide optional spread-rate priors for sensitivity analysis. Third, an RF-informed CA combines the RF probability layer with neighbourhood-driven spread on a 5 km grid. The RF model achieved AUC values of 0.725–0.795 on the 2022–2024 datasets, while the RF-informed CA achieved substantially higher spatial overlap than the evaluated CA-only baselines in the 2023 simulation. A higher-resolution simulation provides an additional qualitative assessment of local spatial errors. These results suggest that combining RF-derived probabilities with local CA spread can improve large-scale wildfire simulations under the tested conditions.

Floor response spectra (FRS) are essential tools for the design of non-structural elements (such as equipment or components). Given the various physical phenomena influencing FRS, high-fidelity (HF) mechanical models of the primary structure may be required to estimate them. Since numerical simulations based on such models are generally computationally expensive, this paper proposes using a multi-fidelity Monte Carlo (MFMC) approach for the efficient estimation of FRS. The method relies on using observations from a fast low-fidelity (LF) model as control variables. If the absolute value of the correlation between LF and HF samples is close to 1, this approach reduces both variance and estimation error compared to a standard Monte Carlo estimate based solely on HF data samples. Through a case study involving the reactor building of the Kashiwazaki-Kariwa nuclear power plant, we demonstrate the suitability of this method for FRS estimation. It effectively reduces variance and estimation error, even when using a LF model as simple as a single-degree-of-freedom system. We also show that the method accounts for modeling uncertainties while maintaining comparable performance. Its ease of use makes it a valuable tool for practitioners.

Classical chaining controls an indexed stochastic process through a single worst-case bound and can therefore obscure substantial variation across the index set. We develop the first simultaneous pointwise majorization theory for Banach-valued processes with finite-metric mixed-tail increments. Suppose that an anchored process $(Z_t)_{t\in T}$ satisfies, for some integer $m\ge1$, pseudo-metrics $d_1,\ldots,d_m$, and orders $\alpha_1,\ldots,\alpha_m>0$, \begin{align*} \mathbb{P}\{\|Z_t-Z_s\|>\sum_{j=1}^m u^{1/\alpha_j}d_j(s,t)\}\le 2e^{-u},s,t\in T. \end{align*} For ambient priors $\mu_1,\ldots,\mu_m$, let $v_j(t):=d_j(t,t_0), \Phi_j(t):=\int_0^{4v_j(t)}(\log\frac{1}{\mu_j(B_{d_j}(t,r))})^{1/\alpha_j}dr$. We prove that, $\forall \delta\in(0,1)$, with probability at least $1-\delta$, simultaneously for all $t\in T$, \begin{align*} \|Z_t\|\le C_{m,\boldsymbol\alpha}\sum_{j=1}^m\{\Phi_j(t)+v_j(t)(\log(e/\delta))^{1/\alpha_j}\}. \end{align*} Here $\boldsymbol\alpha:=(\alpha_1,\ldots,\alpha_m)$ and $C_{m,\boldsymbol\alpha}$ depend only on $m$ and these tail orders. The result subsumes single-metric sub-Weibull processes of every positive order as the case $m=1$. In the Gaussian setting, it sharpens the pointwise upper bound of \citet{xu2026} by eliminating the logarithmic terms generated by dyadic peeling. The proof retains the index-wise costs of measure-generated admissible chains and synchronizes the regimes through a nested common refinement. Finally, we apply our theorems to stationary diffusion empirical processes and decoupled Gaussian chaos to obtain simultaneous pointwise envelope bounds, which can further be applied to other statistics problems.

Feature-based newsvendor models use observable covariates to tailor inventory decisions, aiming to balance holding and shortage costs under demand uncertainty. However, high-dimensional feature sets often hinder interpretability and inflate data collection and implementation costs. This paper studies variable selection for the feature-based newsvendor problem under a hard cardinality constraint on the number of selected features. We formulate the resulting $\ell_0$-constrained empirical newsvendor problem with $\ell_2$-regularization, establish its computational hardness, and develop a mixed-integer second-order cone programming reformulation that strengthens the standard Big-$M$ formulation. To enable scalability beyond exact optimization, we develop a randomized-rounding algorithm with a bi-criteria guarantee and a greedy heuristic. Statistically, we provide theoretical analysis of the resulting sparse policy estimator, including finite-sample estimation error, out-of-sample risk bounds, and support recovery guarantees. Extensive experiments on both synthetic and real data illustrate the computational and statistical trade-offs among various baselines. Our results demonstrate that the proposed variable selection framework achieves competitive out-of-sample operational costs while using substantially fewer covariates.

Historically, retrieving rainfall data from satellite imagery has been the domain of space agencies. However, in recent years, the development of cheaper, more compact satellites (SmallSats) capable of detecting rainfall proxies has led to a significant increase in private-sector initiatives for satellite launch and surface precipitation products. This rapid growth has yet to be matched by data validation efforts. Consequently, the need for a robust tool to detect anomalies in near-real-time data before it is disseminated to the public has become critical. In this paper, we present the development of an anomaly-detection system to identify artifacts in global satellite-based rainfall products. The developed framework leverages pre-trained computer vision models and incorporates scarce human-labeled data to detect specific anomalies. Our proposed anomaly detection strategy is tested on data from the Special Sensor Microwave Imager (SSMI) and the Special Sensor Microwave Imager/Sounder (SSMIS). Results demonstrate the efficacy of our approach at separating regular orbits from artifact-containing orbits for each satellite, with performance comparable to state-of-the-art in-place methods. Additionally, the framework offers explainability and the capacity for iterative refinement following false-positive or false-negative classifications.

We introduce a statistical framework for Gaussian fields indexed at arbitrary edge locations on general compact directed metric graphs. The construction is based on a stochastic differential equation with a first-order operator and conditions at the vertices. We characterise well-posedness and identify the covariance reproducing kernel Hilbert space. We also connect the proposed framework to earlier stream-network models, showing that these arise from the same system under particular boundary conditions, and introduce new boundary conditions that yield more physically realistic processes. The differential-equation representation enables computationally efficient inference and prediction. This makes the method applicable to large data sets without approximation. Applications to temperature modelling on river networks and traffic speeds on road networks illustrate the framework, including the computational efficiency and improved performance under physically informed vertex conditions.

Generative data augmentation is widely used to mitigate class imbalance, yet its theoretical effect on downstream generalization remains poorly understood. In this work, we develop a statistical framework for conditional generative augmentation and analyze its impact on classification risk. We formalize augmentation as a distribution-mixing process and show that the resulting risk distortion is controlled by both the augmentation strength and the class-conditional Wasserstein discrepancy between real and generated distributions. We further derive a capacity-dependent generalization bound based on Rademacher complexity, revealing an explicit trade-off between hypothesis complexity, augmentation intensity, and generative fidelity. Empirically, we evaluate the framework on binary and multiclass imbalanced classification tasks using Conditional GAN and Conditional WGAN-GP augmentation. Across datasets, CWGAN-GP consistently achieves lower Wasserstein discrepancies than CGAN, indicating improved distributional fidelity. However, improved fidelity does not necessarily translate into superior classification performance, with classical oversampling methods often remaining competitive. These findings support the central theoretical prediction that augmentation reliability is governed by distributional approximation error rather than predictive performance alone. Overall, this work establishes generative augmentation as a distributional perturbation process whose reliability can be quantified through Wasserstein-based measures and supported by finite-sample generalization guarantees. The proposed framework provides a principled foundation for evaluating synthetic data quality beyond classification accuracy alone.

The Rashomon effect is a machine learning phenomenon where equally accurate models produce different predictions for the same inputs (predictive multiplicity). Existing work primarily focuses on multiplicity within individual models, but in more complex decision systems, the impact of the Rashomon effect is less well understood. In this work, we study multiplicity from the perspective of auditing incorrect ensemble predictions, where the decision to divert an instance for human review is based on a consistency criterion that combines the ensemble margin with a measure of local prediction variability for each constituent model. With mild assumptions about stability and smoothness, we show that the consistency scores of finite ensembles converge to the corresponding consistency score of the expected model from the Rashomon set as the ensemble size and the number of samples used to measure local prediction variability increase. To demonstrate the efficacy of the proposed criterion, we evaluate the framework with respect to transformer models applied to natural language understanding tasks and parameter-efficient fine-tuning of large language models used for tabular data classification tasks. Our experiments show that ensembling models from the Rashomon set substantially reduces the risk of incorrect predictions going unchecked compared with auditing a single model, while incurring only a moderate increase in the number of diversions. Moreover, the auditing behavior of the full Rashomon set can be closely approximated by finite ensembles of relatively modest size, with the risk approaching zero for some datasets. We further demonstrate that the proposed measure exhibits stronger agreement with established predictive multiplicity metrics than existing consistency measures, providing a more reliable way to capture multiplicity in the Rashomon set.

Persistence diagrams summarize the multiscale topological structure of data, and in applications they often arrive paired with covariates. We develop nonparametric methodology and theory for estimating the expected weighted persistence diagram conditional on a Euclidean covariate. Representing each weighted diagram as a finite random measure on a compact window, we take the density of its conditional expectation as the regression target, the conditional weighted persistence intensity. For a conditional double-kernel estimator we establish finite-sample sup-norm rates with a matching minimax lower bound, uniform rates in partial optimal transport, and an unbiased-risk cross-validation criterion for bandwidth selection. Simulations with analytically known intensities corroborate the theory and show that cross-validation selects the oracle candidate bandwidth in the exact-intensity design. The method is illustrated by studying how radial geometry in cerebral artery trees varies with age.

At a junction, a score field can reveal weighted tangent rays, yet these first-order quantities do not determine how individual branches bend or how their densities change away from the center. Recovering this missing information is necessary for describing local continuation beyond a single point, but finite observations must separate branchwise second-order effects while allowing error in the estimated center. We address this inverse problem using matched score queries at noise scales $\sigma$ and $\lambda\sigma$. For a finite union of $C^{2,\alpha}$ half-branches in $\mathbb{R}^D$, the normalized score has the expansion $F_\sigma=F_0+\sigma G+O(\sigma^{1+\alpha})$. Matched subtraction cancels the tangent contribution and exposes $G$, which depends linearly on branchwise curvature and log-density slope. Given tangent directions and weights on distinct rays, $G$ uniquely identifies all $sD$ branch parameters, and $sD$ scalar component observations are necessary. An $O(\sigma^2)$ center error introduces $D$ translation modes, leading to $(s+1)D$ observations under full-rank calibration, except for a translation-invariant full line. We also establish a perturbation bound and a conditional kernel-density-estimation rate. Experiments reproduce the predicted population and $N^{-1/5}$ trends and remain full rank up to $D=20$ with 16 supplied branches. In end-to-end tests for $D=3$–$5$, a known-count first-order frontend yields full rank in all 135 population systems and a median relative jet error of 0.132. With strong first-order error, matched responses reduce median parameter error by a factor of 49.4 relative to naive tangent subtraction.

We extend recent work establishing an equivalence between one-layer transformers and nearest-neighbor classifiers in the binary setting to the multiclass case. By leveraging the simplex encoding, we show that one-layer transformers with an argmax classification head behave identically to a one-nearest-neighbor classifier in the multiclass setting. This closes a gap left by prior work, whose multiclass result relied on a non-standard rounding-based approach rather than the typical argmax head used in practice.

We prove that a multi-head scaled dot product attention can be viewed as a parameter identification strategy. The ratio of unidentified parameters to the total number of parameters scales like the reciprocal of the number of heads ($1/2 \to 1/(2H)$), meaning models with more heads are structurally more identified. A subtle side effect of the mathematics observation that attention can never be fully identified. Similarly we also show that some bias terms can have no effect on softmax-based attention layers in both the single- and multiple-head settings, though this is mostly a curiosity that should have a marginal effect on model size and model training/prediction efficiency. We also touch on modern improvements to transformers including RoPE and GQA from this perspective, illustrating how those as well can improve the ratio of “meaningful” parameters to all parameters. Simple numerical examples demonstrate that training can indeed involve updates that overlap model-invariant subspaces that arise from a lack of identification. As part of our experiments we use a “rebalancing” approach that can “fix” updates that overlap unindentified subspaces but do not try to present evidence this should actually be adopted. Instead we simply view our numerical results as exploring and confirming the theoretical results. As a whole we discuss a purely mathematical/statistical explanation, identification, for why specific architectural choices in transformers may have improved performance.

The present research is devoted to the nonparametric estimation of a density-dependent drift coefficient in a multivariate McKean–Vlasov diffusion from independent observations at a common time, as well as the stationary density. Under certain assumptions on the (known) potential, we reduce the problem to the one-dimensional one and construct a sieve maximum-likelihood estimator based on sparse ReQU neural networks subject to structural and Hölder constraints. Using the endpoint-adapted graded approximation, we achieve the rate of $\left(b_n\log n/n\right)^{2(\beta+1)/(2\beta+3)}$ for the Kullback-Leibler divergence between the true and estimated stationary densities, with $b_n$ being at most a logarithmic factor. Similarly, it is shown that the constructed estimator for the drift coefficient converges to the true one at the rate of $\left(b_n\log n/n\right)^{\beta/(2\beta+3)}$ in the $L^2$-metric. A matching Assouad lower bound proves minimax optimality of this bound up to logarithmic factors.

Multinomial probit choice probabilities over n alternatives are Gaussian orthant integrals, computed by simulation for thirty years, one expensive integral per alternative. Inversion, which is to say determining item attractiveness consistent with a prescribed choice probability vector, is even more difficult and has been considered impractical for correlated contests when n is large. Yet here, for families lying within a grammar including factor, block and hierarchical covariance structures, we exhibit a calibration tested at n = 1,000,000 reproducing probabilities to very high accuracy, even in the extreme tail. We must return to much smaller problems for any performance comparison to be possible due to limitations of the prior art. The Geweke-Hajivassiliou-Keane simulator is the standard (and still appropriate for high rank) but is two hundred times slower already at n = 200, and its measured cost grows roughly as n^2.8 while ours is linear. Furthermore our approach applies to any continuous performance distributions within reason: the Thurstone-Mosteller model families thereby become a practical alternative to logit at modern scale.