Papers

We propose an approach for dynamic efficiency evaluation across multiple organizational dimensions using data envelopment analysis (DEA). The method generates both dimension-specific and aggregate efficiency scores, incorporates desirable and undesirable outputs, and is suitable for large-scale problem settings. Two regularized DEA models are introduced: a slack-based measure (SBM) and a linearized version of a nonlinear goal programming model (GP-SBM). While SBM estimates an aggregate efficiency score and then distributes it across dimensions, GP-SBM first estimates dimension-level efficiencies and then derives an aggregate score. Both models utilize a regularization parameter to enhance discriminatory power while also directly integrating both desirable and undesirable outputs. We demonstrate the computational efficiency and validity of our approach on multiple datasets and apply it to a case study of twelve hospitals in Ontario, Canada, evaluating three theoretically grounded dimensions of organizational effectiveness over a 24-month period from January 2018 to December 2019: technical efficiency, clinical efficiency, and patient experience. Our numerical results show that SBM and GP-SBM better capture correlations among input/output variables and outperform conventional benchmarking methods that separately evaluate dimensions before aggregation.

Vine copulas offer flexible multivariate dependence modeling and have become widely used in machine learning. Yet, structure learning remains a key challenge. Early heuristics, such as Dissmann's greedy algorithm, are still considered the gold standard but are often suboptimal. We propose random search algorithms and a statistical framework based on model confidence sets, to improve structure selection, provide theoretical guarantees on selection probabilities and excess risk, as well as serve as a foundation for ensembling. Empirical results on real-world data sets show that our methods consistently outperform state-of-the-art approaches.

Tze Leung Lai made seminal contributions to sequential analysis, particularly in sequential hypothesis testing, changepoint detection and nonlinear renewal theory. His work established fundamental optimality results for the sequential probability ratio test and its extensions, and provided a general framework for testing composite hypotheses. In changepoint detection, he introduced new optimality criteria and computationally efficient procedures that remain influential. He applied these and related tools to problems in biostatistics. In this article, we review these key results in the broader context of sequential analysis.

This study develops an AI-based pose estimation pipeline for quantifying movement kinematics in resistance training. Using videos from Wolf et al. (2025), comprising 303 recordings of 26 participants performing eight upper-body exercises under full (fROM) and lengthened partial (pROM) conditions, we extract joint-angle trajectories using five distinct deep-learning pose estimation models and a unified signal-processing framework. From these trajectories, we derive repetition-level metrics including range of motion (ROM) and repetition duration. We use these outputs as dependent variables in a crossed random-effects model that accounts for participant-, exercise-, and model-level variability to assess systematic differences between ROM conditions. Results indicate that pROM reduces range of motion without significantly affecting repetition duration. Variance decomposition shows that pROM increases both between-participant and between-exercise variability, suggesting reduced consistency in execution. To enable cross-exercise comparison, we model ROM on a logarithmic scale and define %ROM as the proportion of fROM achieved under pROM. While the estimated mean is approximately 56%, significant heterogeneity across exercises indicates that lengthened partials are not characterized by a fixed proportion of full ROM. The results demonstrate that AI-based motion analysis can provide reliable kinematic insights to inform evidence-based training recommendations.

We present a novel enhanced cyclic coordinate descent (ECCD) framework for solving generalized linear models with elastic net constraints that reduces training time in comparison to existing state-of-the-art methods. We redesign the CD method by performing a Taylor expansion around the current iterate to avoid nonlinear operations arising in the gradient computation. By introducing this approximation, we are able to unroll the vector recurrences occurring in the CD method and reformulate the resulting computations into more efficient batched computations. We show empirically that the recurrence can be unrolled by a tunable integer parameter, $s$, such that $s > 1$ yields performance improvements without affecting convergence, whereas $s = 1$ yields the original CD method. A key advantage of ECCD is that it avoids the convergence delay and numerical instability exhibited by block coordinate descent. Finally, we implement our proposed method in C++ using Eigen to accelerate linear algebra computations. Comparison of our method against existing state-of-the-art solvers shows consistent performance improvements of $3\times$ in average for regularization path variant on diverse benchmark datasets. Our implementation is available at https://github.com/Yixiao-Wang-Stats/ECCD.

This paper presents a unified geometric framework for Brownian motion on manifolds, encompassing intrinsic Riemannian manifolds, embedded submanifolds, and Lie groups. The approach constructs the stochastic differential equation by injecting noise along each axis of an orthonormal frame and designing the drift term so that the resulting generator coincides with the Laplace–Beltrami operator. Both Stratonovich and Itô formulations are derived explicitly, revealing the geometric origin of curvature-induced drift. The drift is shown to correspond to the covariant derivatives of the frame fields for intrinsic manifolds, the mean curvature vector for embedded manifolds, and the adjoint-trace term for Lie groups, which vanishes for unimodular cases. The proposed formulation provides a geometrically transparent and mathematically consistent foundation for diffusion processes on nonlinear configuration spaces.

Kernel Density Estimation (KDE) is a cornerstone of nonparametric statistics, yet it remains sensitive to bandwidth choice, boundary bias, and computational inefficiency. This study revisits KDE through a principled convolutional framework, providing an intuitive model-based derivation that naturally extends to constrained domains, such as positive-valued random variables. Building on this perspective, we introduce SHIDE (Simulation and Histogram Interpolation for Density Estimation), a novel and computationally efficient density estimator that generates pseudo-data by adding bounded noise to observations and applies spline interpolation to the resulting histogram. The noise is sampled from a class of bounded polynomial kernel densities, constructed through convolutions of uniform distributions, with a natural bandwidth parameter defined by the kernel's support bound. We establish the theoretical properties of SHIDE, including pointwise consistency, bias-variance decomposition, and asymptotic MISE, showing that SHIDE attains the classical $n^{-4/5}$ convergence rate while mitigating boundary bias. Two data-driven bandwidth selection methods are developed, an AMISE-optimal rule and a percentile-based alternative, which are shown to be asymptotically equivalent. Extensive simulations demonstrate that SHIDE performs comparably to or surpasses KDE across a broad range of models, with particular advantages for bounded and heavy-tailed distributions. These results highlight SHIDE as a theoretically grounded and practically robust alternative to traditional KDE.

Differentially private (DP) decentralized Federated Learning (FL) allows local users to collaborate without sharing their data with a central server. However, accurately quantifying the privacy budget of private FL algorithms is challenging due to the co-existence of complex algorithmic components such as decentralized communication and local updates. This paper addresses privacy accounting for two decentralized FL algorithms within the $f$-differential privacy ($f$-DP) framework. We develop two new $f$-DP-based accounting methods tailored to decentralized settings: Pairwise Network $f$-DP (PN-$f$-DP), which quantifies privacy leakage between user pairs under random-walk communication, and Secret-based $f$-Local DP (Sec-$f$-LDP), which supports structured noise injection via shared secrets. By combining tools from $f$-DP theory and Markov chain concentration, our accounting framework captures privacy amplification arising from sparse communication, local iterations, and correlated noise. Experiments on synthetic and real datasets demonstrate that our methods yield consistently tighter $(\epsilon,\delta)$ bounds and improved utility compared to Rényi DP-based approaches, illustrating the benefits of $f$-DP in decentralized privacy accounting.

Multi-omics data integration is crucial for understanding complex diseases, yet limited sample sizes, noise, and heterogeneity often reduce predictive power. To address these challenges, we introduce Omics-GAN, a Generative Adversarial Network (GAN)-based framework designed to generate high-quality synthetic multi-omics profiles while preserving biological relationships. We evaluated Omics-GAN on three omics types (mRNA, miRNA, and DNA methylation) using the ROSMAP cohort for Alzheimer's disease (AD) and TCGA datasets for colon and liver cancer. A support vector machine (SVM) classifier with repeated 5-fold cross-validation demonstrated that synthetic datasets consistently improved prediction accuracy compared to original omics profiles. The AUC of SVM for mRNA improved from 0.72 to 0.74 in AD, and from 0.68 to 0.72 in liver cancer. Synthetic miRNA enhanced classification in colon cancer from 0.59 to 0.69, while synthetic methylation data improved performance in liver cancer from 0.64 to 0.71. Boxplot analyses confirmed that synthetic data preserved statistical distributions while reducing noise and outliers. Feature selection identified significant genes overlapping with original datasets and revealed additional candidates validated by GO and KEGG enrichment analyses. Finally, molecular docking highlighted potential drug repurposing candidates, including Nilotinib for AD, Atovaquone for liver cancer, and Tecovirimat for colon cancer. Omics-GAN enhances disease prediction, preserves biological fidelity, and accelerates biomarker and drug discovery, offering a scalable strategy for precision medicine applications.

As a core policy tool for China in addressing climate risks, green finance plays a strategically important role in shaping carbon mitigation outcomes. This study investigates the nonlinear and interaction effects of green finance on carbon emission intensity (CEI) using Chinese provincial panel data from 2000 to 2022. The Climate Physical Risk Index (CPRI) is incorporated into the analytical framework to assess its potential role in shaping carbon outcomes. We employ Bayesian Additive Regression Trees (BART) to capture complex nonlinear relationships and interaction pathways, and use SHapley Additive exPlanations values to enhance model interpretability. Results show that the Green Finance Index (GFI) has a statistically significant inverted U-shaped effect on CEI, with notable regional heterogeneity. Contrary to expectations, CPRI does not show a significant impact on carbon emissions. Further analysis reveals that in high energy consumption scenarios, stronger green finance development contributes to lower CEI. These findings highlight the potential of green finance as an effective instrument for carbon intensity reduction, especially in energy-intensive contexts, and underscore the importance of accounting for nonlinear effects and regional disparities when designing and implementing green financial policies.

Stochastic Gradient Descent (SGD) has become a cornerstone method in modern data science. However, deploying SGD in high-stakes applications necessitates rigorous quantification of its inherent uncertainty. In this work, we establish non-asymptotic Berry–Esseen bounds for linear functionals of online least-squares SGD, thereby providing a Gaussian Central Limit Theorem (CLT) in a growing-dimensional regime. Existing approaches to high-dimensional inference for projection parameters, such as~\cite{chang2023inference}, rely on inverting empirical covariance matrices and require at least $t \gtrsim d^{3/2}$ iterations to achieve finite-sample Berry–Esseen guarantees, rendering them computationally expensive and restrictive in the allowable dimensional scaling. In contrast, we show that a CLT holds for SGD iterates when the number of iterations grows as $t \gtrsim d^{1+\delta}$ for any $\delta > 0$, significantly extending the dimensional regime permitted by prior works while improving computational efficiency. The proposed online SGD-based procedure operates in $\mathcal{O}(td)$ time and requires only $\mathcal{O}(d)$ memory, in contrast to the $\mathcal{O}(td^2 + d^3)$ runtime of covariance-inversion methods. To render the theory practically applicable, we further develop an online variance estimator for the asymptotic variance appearing in the CLT and establish high-probability deviation bounds for this estimator. Collectively, these results yield the first fully online and data-driven framework for constructing confidence intervals for SGD iterates in the near-optimal scaling regime $t \gtrsim d^{1+\delta}$.

We propose correlated semi-implicit variational inference (Co-SIVI), a scalable approach for full posterior approximation in large spatial models with exponential-family likelihoods. Co-SIVI incorporates dependence directly into the conditional variational distribution of spatial random effects through an iterative weighted least squares algorithm that accommodates both Gaussian process and nearest-neighbor Gaussian process (NNGP) priors. For large samples, we further propose reparameterizing the variational family for the covariance parameters to better capture posterior dependence. Co-SIVI addresses an important limitation of semi-implicit variational inference (SIVI), for which dependence induced through the mixing distribution may be insufficient when spatial random effects are explicitly included in the variational family. In simulations with Gaussian, Poisson, Gamma, and Bernoulli outcomes, Co-SIVI closely reproduces Hamiltonian Monte Carlo (HMC) results at substantially lower computational cost, while SIVI performs similarly for marginalized Gaussian models. We apply SIVI and Co-SIVI, both with NNGP priors, to two large-scale datasets: temperature data modeled with a Gaussian likelihood and housing price data modeled with a Gamma likelihood. Overall, Co-SIVI provides a scalable and flexible alternative to HMC for full posterior approximation in large non-Gaussian spatial models.

Policy learning algorithms are widely used in areas such as personalized medicine and advertising to develop individualized treatment regimes. However, most methods force a decision even when predictions are uncertain, which is risky in high-stakes settings. We study policy learning with abstention, where a policy may defer to a safe default or an expert. When a policy abstains, it receives a small additive reward on top of the value of a random guess. We propose a two-stage learner that first identifies a set of near-optimal policies and then constructs an abstention rule from their disagreements. We establish fast O(1/n)-type regret guarantees when propensities are known, and extend these guarantees to the unknown-propensity case via a doubly robust (DR) objective. We further show that abstention is a versatile tool with direct applications to other core problems in policy learning: it yields improved guarantees under margin conditions without the common realizability assumption, connects to distributionally robust policy learning by hedging against small data shifts, and supports safe policy improvement by ensuring improvement over a baseline policy with high probability.

This paper develops a continuous functional framework for analyzing contagion dynamics in financial networks, extending the Navier-Stokes-based approach to network-structured spatial processes. We model financial distress propagation as a diffusion process on weighted networks, deriving a network diffusion equation from first principles that predicts contagion decay depends on the network's algebraic connectivity through the relation $\kappa = \sqrt{\lambda_2/D}$, where $\lambda_2$ is the second-smallest eigenvalue of the graph Laplacian and $D$ is the diffusion coefficient. Applying this framework to European banking data from the EBA stress tests (2018, 2021, 2023), we estimate interbank exposure networks using maximum entropy methods and track the evolution of systemic risk through the COVID-19 crisis. Our key finding is that network connectivity declined by 45% from 2018 to 2023, implying a 26% reduction in the contagion decay parameter. Difference-in-differences analysis reveals this structural change was driven by regulatory-induced deleveraging of systemically important banks, which experienced differential asset reductions of 17% relative to smaller institutions. The networks exhibit lognormal rather than scale-free degree distributions, suggesting greater resilience than previously assumed in the literature. Extensive robustness checks across parametric and non-parametric estimation methods confirm declining systemic risk, with cross-method correlations exceeding 0.95. These findings demonstrate that post-COVID-19 regulatory reforms effectively reduced network interconnectedness and systemic vulnerability in the European banking system.

Repeated multi-unit auctions, where a seller allocates multiple identical items over many rounds, are common mechanisms in electricity markets and treasury auctions. We compare the two predominant formats: uniform-price and discriminatory auctions, focusing on the perspective of a single bidder learning to bid against stochastic adversaries. We characterize the learning difficulty in each format, showing that the regret scales similarly for both auction formats under both full-information and bandit feedback, as $\tilde{\Theta} ( \sqrt{T} )$ and $\tilde{\Theta} ( T^{2/3} )$, respectively. However, analysis beyond worst-case regret reveals structural differences: uniform-price auctions may admit faster learning rates, with regret scaling as $\tilde{\Theta} ( \sqrt{T} )$ in settings where discriminatory auctions remain at $\tilde{\Theta} ( T^{2/3} )$. Finally, we provide a specific analysis for auctions in which the other participants are symmetric and have unit-demand, and show that in these instances, a similar regret rate separation appears.

Estimating the true rank of a noisy data matrix is a fundamental problem underlying techniques such as principal component analysis, matrix completion, etc. Existing rank estimation criteria, including information-based and cross-validation methods, are either highly sensitive to outliers or computationally demanding when combined with robust estimators. This paper proposes a new criterion, the Divergence Information Criterion for Matrix Rank (DICMR), that achieves both robustness and computational simplicity. Derived from the density power divergence framework, DICMR inherits the robustness properties while being computationally very simple. We provide asymptotic bounds on its overestimation and underestimation probabilities, and demonstrate first-order B-robustness of the criteria. Extensive simulations show that DICMR delivers accuracy comparable to the robustified cross-validation methods, but with far lower computational cost. We also showcase a real-data application to microarray imputation to further demonstrate its practical utility, outperforming several state-of-the-art algorithms.

We investigate the fundamental problem of leveraging offline data to accelerate online reinforcement learning - a direction with strong potential but limited theoretical grounding. Our study centers on how to learn and apply value envelopes within this context. To this end, we introduce a principled two-stage framework: the first stage uses offline data to derive upper and lower bounds on value functions, while the second incorporates these learned bounds into online algorithms. Our method extends prior work by decoupling the upper and lower bounds, enabling more flexible and tighter approximations. In contrast to approaches that rely on fixed shaping functions, our envelopes are data-driven and explicitly modeled as random variables, with a filtration argument ensuring independence across phases. The analysis establishes high-probability regret bounds determined by two interpretable quantities, thereby providing a formal bridge between offline pre-training and online fine-tuning. Empirical results on tabular MDPs demonstrate substantial regret reductions compared with both UCBVI and prior methods while remaining competitive with related approaches.

In probabilistic principal component analysis (PPCA), an observed vector is modeled as a linear transformation of a low-dimensional Gaussian factor plus isotropic noise. We generalize PPCA to tensors by constraining the loading operator to have Tucker structure, yielding a probabilistic multilinear PCA model that enables uncertainty quantification and naturally accommodates multiple, possibly heterogeneous, tensor observations. We develop the associated theory: we establish identifiability of the loadings and noise variance and show that-unlike in matrix PPCA-the maximum likelihood estimator (MLE) exists even from a single tensor sample. We then study two estimators. First, we consider the MLE and propose an expectation maximization (EM) algorithm to compute it. Second, exploiting that Tucker maps correspond to rank-one elements after a Kronecker lifting, we design a computationally efficient estimator for which we provide provable finite-sample guarantees. Together, these results provide a coherent probabilistic framework and practical algorithms for learning from tensor-valued data.

Simulation studies are widely used to evaluate statistical methods. However, new methods are often introduced and evaluated using data-generating mechanisms (DGMs) devised by the same authors. This coupling creates misaligned incentives, e.g., the need to demonstrate the superiority of new methods, potentially compromising the neutrality of simulation studies. Furthermore, results of simulation studies are often difficult to compare due to differences in DGMs, competing methods, and performance measures. This fragmentation can lead to conflicting conclusions, hinder methodological progress, and delay the adoption of effective methods. To address these challenges, we introduce the concept of living synthetic benchmarks. The key idea is to disentangle method and simulation study development and continuously update the benchmark whenever a new DGM, method, or performance measure becomes available. This separation benefits the neutrality of method evaluation, emphasizes the development of both methods and DGMs, and enables systematic comparisons. In this paper, we outline a blueprint for building and maintaining such benchmarks, discuss the technical and organizational challenges of implementation, and demonstrate feasibility with a prototype benchmark for publication bias adjustment methods. We conclude that living synthetic benchmarks have the potential to foster neutral, reproducible, and cumulative evaluation of methods, benefiting both method developers and users.

Understanding the dynamics of feature learning in neural networks (NNs) remains a significant challenge. The work of (Mousavi-Hosseini et al., 2023) analyzes a multiple index teacher-student setting and shows that a two-layer student attains a low-rank structure in its first-layer weights when trained with stochastic gradient descent (SGD) and a strong regularizer. This structural property is known to reduce sample complexity of generalization. Indeed, in a second step, the same authors establish algorithm-specific learning guarantees under additional assumptions. In this paper, we focus exclusively on the structure discovery aspect and study it under weaker assumptions, more specifically: we allow (a) NNs of arbitrary size and depth, (b) with all parameters trainable, (c) under any smooth loss function, (d) tiny regularization, and (e) trained by any method that attains a second-order stationary point (SOSP), e.g.\ perturbed gradient descent (PGD). At the core of our approach is a key $\textit{derandomization}$ lemma, which states that optimizing the function $\mathbb{E}_{\mathbf{x}} \left[g_{\theta}(\mathbf{W}\mathbf{x} + \mathbf{b})\right]$ converges to a point where $\mathbf{W} = \mathbf{0}$, under mild conditions. The fundamental nature of this lemma directly explains structure discovery and has immediate applications in other domains including an end-to-end approximation for MAXCUT, and computing Johnson-Lindenstrauss embeddings.

We revisit Cox's proportional hazards model to improve variable selection in survival analysis. A square-root transformation of the partial likelihood renders the selection of the regularization parameter pivotal, free of the unknown baseline hazard and censoring mechanism. The resulting criterion borrows from information criteria such as BIC and from penalized regression methods such as the lasso, taking the best of both. On simulated and real data, our method substantially improves upon state-of-the-art approaches used daily in support recovery.

We study reinforcement learning (RL) with transition look-ahead, where the agent may observe which states would be visited upon playing any sequence of $\ell$ actions before deciding its course of action. While such predictive information can drastically improve the achievable performance, we show that using this information optimally comes at a potentially prohibitive computational cost. Specifically, we prove that optimal planning with one-step look-ahead ($\ell=1$) can be solved in polynomial time through a novel linear programming formulation. In contrast, for $\ell \geq 2$, the problem becomes NP-hard. Our results delineate a precise boundary between tractable and intractable cases for the problem of planning with transition look-ahead in reinforcement learning.

In this paper, we introduce PSI-LinUCB, a scalable variant of LinUCB that enables efficient training, inference, and memory usage by representing the inverse regularized design matrix as a sum of a diagonal matrix and low-rank correction. We derive numerically stable rank-1 and batched updates that maintain the inverse without explicitly forming the matrix. To control memory growth, we employ a projector-splitting integrator for dynamical low-rank approximation, yielding an average per-step update cost and memory usage of $O(dr)$ for approximation rank $r$. The inference complexity of the proposed algorithm is $O(dr)$ per action evaluation. Experiments on recommender system datasets demonstrate the effectiveness of our algorithm.

The advent of Large Language Models has revolutionized tasks across domains, including the automation of legal document analysis, a critical component of modern contract management systems. This paper presents a comprehensive implementation of LLM-enhanced metadata extraction for contract review, focusing on the automatic detection and annotation of salient legal clauses. Leveraging both the publicly available Contract Understanding Atticus Dataset (CUAD) and proprietary contract datasets, our work demonstrates the integration of advanced LLM methodologies with practical applications. We identify three pivotal elements for optimizing metadata extraction: robust text conversion, strategic chunk selection, and advanced LLM-specific techniques, including Chain of Thought (CoT) prompting and structured tool calling. The results from our experiments highlight the substantial improvements in clause identification accuracy and efficiency. Our approach shows promise in reducing the time and cost associated with contract review while maintaining high accuracy in legal clause identification. The results suggest that carefully optimized LLM systems could serve as valuable tools for legal professionals, potentially increasing access to efficient contract review services for organizations of all sizes.

This paper deals with the GMANOVA model with a matrix of polynomial basis functions as a within-individual design matrix. The model involves two model selection problems: the selection of explanatory variables and the selection of the degrees of the polynomials. The two problems can be uniformly addressed by hierarchically incorporating zeros into the vectors of regression coefficients. Based on this idea, we propose hierarchical overlapping group Lasso (HOGL) to perform the variable and degree selections simultaneously. Importantly, when using a polynomial basis, fitting a highdegree polynomial often causes problems in model selection. In the approach proposed here, these problems are handled by using a matrix of orthonormal basis functions obtained by transforming the matrix of polynomial basis functions. Algorithms are developed with optimality and convergence to optimize the method. The performance of the proposed method is evaluated using numerical simulation.

This study investigates whether Topological Data Analysis (TDA) can provide additional insights beyond traditional statistical methods in clustering currency behaviours. We focus on the foreign exchange (FX) market, which is a complex system often exhibiting non-linear and high-dimensional dynamics that classical techniques may not fully capture. We compare clustering results based on TDA-derived features versus classical statistical features using monthly logarithmic returns of 13 major currency exchange rates (all against the euro). Two widely-used clustering algorithms, \(k\)-means and Hierarchical clustering, are applied on both types of features, and cluster quality is evaluated via the Silhouette score and the Calinski-Harabasz index. Our findings show that TDA-based feature clustering produces more compact and well-separated clusters than clustering on traditional statistical features, particularly achieving substantially higher Calinski-Harabasz scores. However, all clustering approaches yield modest Silhouette scores, underscoring the inherent difficulty of grouping FX time series. The differing cluster compositions under TDA vs. classical features suggest that TDA captures structural patterns in currency co-movements that conventional methods might overlook. These results highlight TDA as a valuable complementary tool for analysing financial time series, with potential applications in risk management where understanding structural co-movements is crucial.

Deep ensembles deliver state-of-the-art, reliable uncertainty quantification, but their heavy computational and memory requirements hinder their practical deployments to real applications such as on-device AI. Knowledge distillation compresses an ensemble into small student models, but existing techniques struggle to preserve uncertainty partly because reducing the size of DNNs typically results in variation reduction. To resolve this limitation, we introduce a new method of distribution distillation (i.e. compressing a teacher ensemble into a student distribution instead of a student ensemble) called Gaussian distillation, which estimates the distribution of a teacher ensemble through a special Gaussian process called the deep latent factor model (DLF) by treating each member of the teacher ensemble as a realization of a certain stochastic process. The mean and covariance functions in the DLF model are estimated stably by using the expectation-maximization (EM) algorithm. By using multiple benchmark datasets, we demonstrate that the proposed Gaussian distillation outperforms existing baselines. In addition, we illustrate that Gaussian distillation works well for fine-tuning of language models and distribution shift problems.

We present a systematic analysis of estimation errors for a class of optimal transport based algorithms for filtering and data assimilation. Along the way, we extend previous error analyses of Brenier maps to the case of conditional Brenier maps that arise in the context of simulation based inference. We then apply these results in a filtering scenario to analyze the optimal transport filtering algorithm of Al-Jarrah et al. (2024, ICML). An extension of that algorithm along with numerical benchmarks on various non-Gaussian and high-dimensional examples are provided to demonstrate its effectiveness and practical potential.

Clustering-based features are widely used in machine learning, but most methods must choose a resolution – a choice that is global, fixed, and ad hoc. Recent work shows that varying the resolution parameter produces only a finite set of structurally stable partitions, known as configurations. Based on this, we introduce Configuration-Mixed Prediction (CMP), a setting where models learn to adaptively weight these configurations per sample for downstream prediction. We propose MixConfig, a plug-and-play feature augmentation module that extracts configurations from any frozen embedding and learns energy-aware mixing weights via a novel selector that jointly reasons about sample context, cluster assignments, and stability statistics. Experiments across tabular, molecular, vision, and text domains demonstrate consistent improvements over single-resolution and static baselines across diverse predictor architectures, with gains particularly pronounced in low-data regimes.

Analyzing data from multiple sources offers valuable opportunities to improve the estimation efficiency of causal estimands. However, this analysis also poses many challenges due to population heterogeneity and data privacy constraints. While several advanced methods for causal inference in federated settings have been developed in recent years, many focus on difference-based averaged causal effects and are not designed to study effect modification. In this study, we introduce a novel targeted-federated learning framework to study the heterogeneity of treatment effects (HTEs) for a targeted population by proposing a projection-based estimand. This HTE framework integrates information from multiple data sources without sharing raw data, while accounting for covariate distribution shifts among sources. Our proposed approach is shown to be doubly robust, conveniently supporting both difference-based estimands for continuous outcomes and odds ratio-based estimands for binary outcomes. Furthermore, we develop a communication-efficient bootstrap-based selection procedure to detect non-transportable data sources, thereby enhancing robust information aggregation without introducing bias. The superior performance of the proposed estimator over existing methods is demonstrated through extensive simulation studies, and the utility of our approach has been shown in a real-world data application using nationwide Medicare-linked data.