Papers

Firms that price perishable resources – airline seats, hotel rooms, seasonal inventory – now routinely use demand predictions, but these predictions vary widely in quality. Under hard capacity constraints, acting on an inaccurate prediction can irreversibly deplete inventory needed for future periods. We study how prediction uncertainty propagates into dynamic pricing decisions with linear demand, stochastic noise, and finite capacity. A certified demand forecast with known error bound~$\epsilon^0$ specifies where the system should operate: it shifts regret from $O(\sqrt{T})$ to $O(\log T)$ when $\epsilon^0 \lesssim T^{-1/4}$, and we prove this threshold is tight. A misspecified surrogate model – biased but correlated with true demand – cannot set prices directly but reduces learning variance by a factor of $(1-\rho^2)$ through control variates. The two mechanisms compose: the forecast determines the regret regime; the surrogate tightens estimation within it. All algorithms rest on a boundary attraction mechanism that stabilizes pricing near degenerate capacity boundaries without requiring non-degeneracy assumptions. Experiments confirm the phase transition threshold, the variance reduction from surrogates, and robustness across problem instances.

This article studies randomization inference for treatment effects in randomized controlled trials with attrition, where outcomes are observed for only a subset of units. We assume monotonicity in reporting behavior as in \cite{lee2009training} and focus on the average treatment effect for always-reporters (AR-ATE), defined as units whose outcomes are observed under both treatment and control. Because always-reporter status is only partially revealed by observed assignment and response patterns, we propose a worst-case randomization test that maximizes the randomization p-value over all always-reporter configurations consistent with the data, with an optional pretest to prune implausible configurations. Using studentized Hajek- and chi-square-type statistics, we show the resulting procedure is finite-sample valid for the sharp null and asymptotically valid for the weak null. We also discuss computational implementations for discrete outcomes and integer-programming-based bounds for continuous outcomes.

Deploying neural networks on microcontrollers is constrained by kilobytes of flash and SRAM, where 1x1 pointwise (PW) mixers often dominate memory even after INT8 quantization across vision, audio, and wearable sensing. We present HYPER-TINYPW, a compression-as-generation approach that replaces most stored PW weights with generated weights: a shared micro-MLP synthesizes PW kernels once at load time from tiny per-layer codes, caches them, and executes them with standard integer operators. This preserves commodity MCU runtimes and adds only a one-off synthesis cost; steady-state latency and energy match INT8 separable CNN baselines. Enforcing a shared latent basis across layers removes cross-layer redundancy, while keeping PW1 in INT8 stabilizes early, morphology-sensitive mixing. We contribute (i) TinyML-faithful packed-byte accounting covering generator, heads/factorization, codes, kept PW1, and backbone; (ii) a unified evaluation with validation-tuned t* and bootstrap confidence intervals; and (iii) a deployability analysis covering integer-only inference and boot versus lazy synthesis. On three ECG benchmarks (Apnea-ECG, PTB-XL, MIT-BIH), HYPER-TINYPW shifts the macro-F1 versus flash Pareto frontier: at about 225 kB it matches a roughly 1.4 MB CNN while being 6.31x smaller (84.15% fewer bytes), retaining at least 95% of large-model macro-F1. Under 32-64 kB budgets it sustains balanced detection where compact baselines degrade. The mechanism applies broadly to other 1D biosignals, on-device speech, and embedded sensing tasks where per-layer redundancy dominates, indicating a wider role for compression-as-generation in resource-constrained ML systems. Beyond ECG, HYPER-TINYPW transfers to TinyML audio: on Speech Commands it reaches 96.2% test accuracy (98.2% best validation), supporting broader applicability to embedded sensing workloads where repeated linear mixers dominate memory.

Bayesian inference over positive semidefinite (PSD) matrix-valued parameters arises in structured covariance estimation, graph-Laplacian precision models, and multi-output graph learning, but Euclidean proposals often mix poorly near the cone boundary. We propose \ConeMALA, a geometry-aware Metropolis-adjusted Langevin algorithm whose proposal geometry is induced by the model's log-determinant structure. For a PSD-weighted graph with edge kernels $W_e\succeq 0$, block Laplacian $L(W)$ , and stabilizer $R\succ 0$, the lifted precision matrix $X(W)=L(W)+R\in \mathbb S_{++}^{md}$ defines the log-determinant energy $\Phi(W)=-\log\det X(W).$ We show that the Hessian of $\Phi$ is the pullback of the affine-invariant SPD metric under the map $W\mapsto X(W)$, yielding explicit intrinsic Langevin proposals with Metropolis-Hastings correction using the closed-form SPD exponential-map Jacobian. We validate the metric on rank-one PSD edge perturbations for $d=5$, obtaining essentially exact agreement between analytic curvature scores and finite-difference curvatures. In intrinsic SPD posterior and matrix-valued graph Gaussian experiments, \ConeMALA achieves stable multichain diagnostics and substantially higher ESS/sec than Euclidean MALA and generic RMALA, while a PDHMC-like finite-difference baseline is accurate but computationally prohibitive at larger graph sizes. These results show that pullback log-determinant geometry provides a practical route to uncertainty quantification in PSD-constrained graph learning.

Community associations rely heavily on resident surveys to guide decisions about amenities, infrastructure, and services. However, survey responses reflect perceptions that may not directly correspond to underlying operational conditions. This study bridges that gap by calibrating survey-based satisfaction measures against objective utilization data. Using parking and facility data from Tellico Village, we map perceived problem rates to utilization exceedance probabilities to estimate behavioral congestion thresholds. Results show that dissatisfaction emerges near effective capacity - once spatial, temporal, and informational constraints are considered - rather than at nominal capacity limits. Perceived difficulty is concentrated among active users and is shaped by operational frictions and incomplete system knowledge. These findings demonstrate that perceived congestion reflects constraints on access and reliability, not simply physical shortages. By distinguishing between effective and nominal capacity, the proposed framework enables more accurate diagnosis of system conditions. We propose incorporating behavioral metrics into community performance frameworks to support better decision-making, reduce unnecessary capital expansion, and target operational improvements more effectively.

Generalized Category Discovery (GCD) leverages labeled data to categorize unlabeled samples from known or unknown classes. Most previous methods jointly optimize supervised and unsupervised objectives and achieve promising results. However, inherent optimization interference still limits their ability to improve further. Through quantitative analysis, we identify a key issue, i.e., gradient entanglement, which 1) distorts supervised gradients and weakens discrimination among known classes, and 2) induces representation-subspace overlap between known and novel classes, reducing the separability of novel categories. To address this issue, we propose the Energy-Aware Gradient Coordinator (EAGC), a plug-and-play gradient-level module that explicitly regulates the optimization process. EAGC comprises two components: Anchor-based Gradient Alignment (AGA) and Energy-aware Elastic Projection (EEP). AGA introduces a reference model to anchor the gradient directions of labeled samples, preserving the discriminative structure of known classes against the interference of unlabeled gradients. EEP softly projects unlabeled gradients onto the complement of the known-class subspace and derives an energy-based coefficient to adaptively scale the projection for each unlabeled sample according to its degree of alignment with the known subspace, thereby reducing subspace overlap without suppressing unlabeled samples that likely belong to known classes. Experiments show that EAGC consistently boosts existing methods and establishes new state-of-the-art results. Code is available at https://haiyangzheng.github.io/EAGC.

In Part I of this series, we established a rigorous mathematical isomorphism between ant colony decision-making and random forest learning, demonstrating that variance reduction through decorrelation is a universal principle shared by biological and computational ensembles. Here we turn to the complementary mechanism: bias reduction through adaptive weighting. Just as boosting algorithms sequentially focus on difficult instances, ant colonies dynamically amplify successful foraging paths through pheromone-mediated recruitment. We prove that these processes are mathematically isomorphic, establishing that the fundamental theorem of weak learnability has a direct analog in colony decision-making. We develop a formal mapping between AdaBoost's adaptive reweighting and ant recruitment dynamics, show that the margin theory of boosting corresponds to the stability of quorum decisions, and demonstrate through comprehensive simulation that ant colonies implementing adaptive recruitment achieve the same bias-reduction benefits as boosting algorithms. This completes a unified theory of ensemble intelligence, revealing that both variance reduction (Part I) and bias reduction (Part II) are manifestations of the same underlying mathematical principles governing collective intelligence in biological and computational systems.

We study the problem of routing queries to large language models (LLMs) under cost, GPU resources, and concurrency constraints. Prior per-query routing methods often fail to control batch-level cost, especially under non-uniform or adversarial batching. To address this, we propose a batch-level, resource-aware routing framework that jointly optimizes model assignment for each batch while respecting cost and model capacity limits. We further introduce a robust variant that accounts for uncertainty in predicted LLM performance, along with an offline instance allocation procedure that balances quality and throughput across multiple models. Experiments on two multi-task LLM benchmarks show that robustness improves accuracy by 1-14% over non-robust counterparts (depending on the performance estimator), batch-level routing outperforms per-query methods by up to 24% under adversarial batching, and optimized instance allocation yields additional gains of up to 3% compared to a non-optimized allocation, all while strictly controlling cost and GPU resource constraints.

Machine learning models of chemical bioactivity are increasingly used for prioritizing a small number of compounds in virtual screening libraries for experimental follow-up. In these applications, assessing model accuracy by early hit enrichment such as Positive Predicted Value (PPV) calculated for top N hits (PPV@N) is more appropriate and actionable than traditional global metrics such as AUC. We present KANEL, an ensemble workflow that combines interpretable Kolmogorov-Arnold Networks (KANs) with XGBoost, random forest, and multilayer perceptron models trained on complementary molecular representations (LillyMol descriptors, RDKit-derived descriptors, and Morgan fingerprints).

We propose a unified framework to draw inferences for regression coefficients in a generalized linear model (GLM) following Lasso-based variable selection. We adapt to non-Gaussian GLMs a recently developed parametric programming strategy for post-selection inference in the linear model with a Gaussian response by drawing parallels between maximum likelihood estimation in GLMs and least squares estimation in linear models. We then conduct post-selection inference based on a linearized model for pseudo response and covariate data strategically created based on the raw data. Using synthetic data generated from regression models for three different types of non-Gaussian responses in simulation experiments, we demonstrate that the proposed method effectively corrects the naive inference that ignores variable selection while achieving greater efficiency than a polyhedral-based post-selection adjustment.

In applications, quantities of interest are often modelled in equilibrium or an equilibrium solution is sought. The presence of confounding makes causal inference in this setting challenging. We provide interpretable graphical models for equilibrium systems with confounding using anterial graphs (Lauritzen and Sadeghi, 2018), a class of graphs containing directed acyclic graphs, ancestral graphs, and chain graphs. In this setting, we provide valid graphical representations of both counterfactual variables and observational variables, which we relate to counterfactual graphs (Shpitser and Pearl, 2007) and single-world intervention graphs (Richardson and Robins,2013). As an application of this graphical representation, we provide an element-wise procedure of selecting adjustment sets that flexibly include and exclude given covariates.

We introduce a flexible framework for high-dimensional matrix estimation to incorporate side information for both rows and columns. Existing approaches, such as inductive matrix completion, often impose restrictive structure-for example, an exact low-rank covariate interaction term, linear covariate effects, and limited ability to exploit components explained only by one side (row or column) or by neither-and frequently omit an explicit noise component. To address these limitations, we propose to decompose the underlying matrix as the sum of four complementary components: (possibly nonlinear) interaction between row and column characteristics; row characteristic-driven component, column characteristic-driven component, and residual low-rank structure unexplained by observed characteristics. By combining sieve-based projection with nuclear-norm penalization, each component can be estimated separately and these estimated components can then be aggregated to yield a final estimate. We derive convergence rates that highlight robustness across a range of model configurations depending on the informativeness of the side information. We further extend the method to partially observed matrices under both missing-at-random and missing-not-at-random mechanisms, including block-missing patterns motivated by causal panel data. Simulations and a real-data application to tobacco sales show that leveraging side information improves imputation accuracy and can enhance treatment-effect estimation relative to standard low-rank and spectral-based alternatives.

This paper introduces robust twoblock (RTB) simultaneous dimension reduction, which is the first statistically robust method to perform simultaneous dimension reduction in two blocks of variables that allows to tune the number of components in each block individually. The paper proposes both a dense and a sparse version of the new method. Sparse RTB is the first robust estimator of this kind that allows to tune both the number of components and the degree of sparsity (i.e.\ the number of variables retained) for each block individually. RTB thereby allows to optimally extract and summarize the relevant portion of information in each block of data, also in the presence of outliers. As a corollary, the estimators can be recombined into a single estimate of regression coefficients for multivariate regression that is operable when the number of variables exceeds the number of cases in each block. An extensive simulation study illustrates that the new methods are resistant to different types of outliers, while maintaining estimation efficiency across a range of dimensionality settings. These findings both hold true for the dense and the sparse method. The methods' performance is further illustrated on two example data sets and a straightforward algorithm is presented and made accessible in an open source repository.

Previous comparisons of ordinary least squares with Newey-West standard errors (OLS-NW) and Prais-Winsten (PW) regression in multiple-group interrupted time series analysis have been limited to first-order autoregressive (AR[1]) errors because PW estimation for higher-order AR[k] processes was previously unavailable. We conducted the first systematic evaluation of OLS-NW and PW under AR[2] and AR[3] error structures using Monte Carlo simulation. Simulations examined mild positive, oscillatory, and high persistent autocorrelation across varying series lengths and effect sizes. OLS-NW generally showed higher apparent power but substantially inflated Type I error and poor coverage, particularly under persistent autocorrelation, where inferential performance worsened with increasing AR order and series length. PW maintained substantially better inferential calibration across nearly all conditions. Both methods were approximately unbiased.

In many scientific applications, hypotheses are generated and tested continuously in a stream. We develop a framework for improving online multiple testing procedures with false discovery rate (FDR) control under arbitrary dependence. Our approach is two-fold: we construct methods via the online e-closure principle, as well as a novel formulation of online compound e-values that is defined through donations. This yields strict power improvements over state-of-the-art e-value and p-value procedures while retaining FDR control. We further derive algorithms that compute the decision at time $t$ in $O(\log t)$ time, and we demonstrate improved empirical performance on synthetic and real data.

It has become standard for empirical studies to conduct inference robust to cluster dependence and heterogeneity. With a small number of clusters, the normal approximation for the $t$-statistics of regression coefficients may be poor. This paper tackles this problem using a critical value based on the conditional Cramér-Edgeworth expansion for the $t$-statistics. The proposed critical value guarantees third-order refinement, and it does not require resampling because it is a closed-form function of the estimated score skewness and kurtosis. Simulations show that our proposal can make a difference in size control with as few as 10 clusters. Keywords: Cluster robust inference, Cramér-Edgeworth expansion, Asymptotic refinement

Causal discovery aims to infer causal relationships among variables from observational data, typically represented by a directed acyclic graph (DAG). Most existing methods assume independent and identically distributed observations, an assumption often violated in modern applications. In addition, many datasets contain a mixture of continuous and discrete variables, which further complicates causal modeling when dependence across samples is present. To address these challenges, we propose a de-correlation framework for causal discovery from dependent mixed data. Our approach formulates a structural equation model with latent variables that accommodates both continuous and discrete variables while allowing correlated Gaussian errors across units. We estimate the dependence structure among samples via a pairwise maximum likelihood estimator for the covariance matrix and develop an EM algorithm to impute latent variables underlying discrete observations. A de-correlation transformation of the recovered latent data enables the use of standard causal discovery algorithms to learn the underlying causal graph. Simulation studies demonstrate that the proposed method substantially improves causal graph recovery compared with applying standard methods directly to the original dependent data. We apply our approach to single-cell RNA sequencing data to infer gene regulatory networks governing embryonic stem cell differentiation. The inferred regulatory networks show significantly improved predictive likelihood on test data, and many high-confidence edges are supported by known regulatory interactions reported in the literature.

Missing data is a ubiquitous challenge in data analysis, often leading to biased and inaccurate results. Traditional imputation methods usually assume that the missingness mechanism is missing-at-random (MAR), where the missingness is independent of the missing values themselves. This assumption is frequently violated in real-world scenarios, prompted by recent advances in imputation methods using deep learning to address this challenge. However, these methods neglect the crucial issue of nonparametric identifiability in missing-not-at-random (MNAR) data, which can lead to biased and unreliable results. This paper seeks to bridge this gap by proposing a novel framework based on deep latent variable models for MNAR data. Building on the assumption of conditional no self-censoring given latent variables, we establish the identifiability of the data distribution. This crucial theoretical result guarantees the feasibility of our approach. To effectively estimate unknown parameters, we develop an efficient algorithm utilizing importance-weighted autoencoders. We demonstrate, both theoretically and empirically, that our estimation process accurately recovers the ground-truth joint distribution under specific regularity conditions. Extensive simulation studies and real-world data experiments showcase the advantages of our proposed method compared to various classical and state-of-the-art approaches to missing data imputation.

Online health communities (OHCs) are vital for fostering peer support and improving health outcomes. Support groups within these platforms can provide more personalized and cohesive peer support, yet traditional support group formation methods face challenges related to scalability, static categorization, and insufficient personalization. To overcome these limitations, we propose two novel machine learning models for automated support group formation: the Group specific Dirichlet Multinomial Regression (gDMR) and the Group specific Structured Topic Model (gSTM). These models integrate user generated textual content, demographic profiles, and interaction data represented through node embeddings derived from user networks to systematically automate personalized, semantically coherent support group formation. We evaluate the models on a large scale dataset from MedHelp, comprising over 2 million user posts. Both models substantially outperform baseline methods including LDA, DMR, and STM in predictive accuracy (held out log likelihood), semantic coherence (UMass metric), and internal group consistency. The gDMR model yields group covariates that facilitate practical implementation by leveraging relational patterns from network structures and demographic data. In contrast, gSTM emphasizes sparsity constraints to generate more distinct and thematically specific groups. Qualitative analysis further validates the alignment between model generated groups and manually coded themes, showing the practical relevance of the models in informing groups that address diverse health concerns such as chronic illness management, diagnostic uncertainty, and mental health. By reducing reliance on manual curation, these frameworks provide scalable solutions that enhance peer interactions within OHCs, with implications for patient engagement, community resilience, and health outcomes.

Conditional independence is central to modern statistics, but beyond special parametric families it rarely admits an exact covariance characterization. We introduce the binary expansion group intersection network (BEGIN), a distribution-free graphical representation for multivariate binary data and bit-encoded multinomial variables. For arbitrary binary random vectors and bit representations of multinomial variables, we prove that conditional independence is equivalent to a sparse linear representation of conditional expectations, to a block factorization of the corresponding interaction covariance matrix, and to block diagonality of an associated generalized Schur complement. The resulting graph is indexed by the intersection of multiplicative groups of binary interactions, yielding an analogue of Gaussian graphical modeling beyond the Gaussian setting. This viewpoint treats data bits as atoms and local BEGIN molecules as building blocks for large Markov random fields. We also show how dyadic bit representations allow BEGIN to approximate conditional independence for general random vectors under mild regularity conditions. A key technical device is the Hadamard prism, a linear map that links interaction covariances to group structure.

In many institutional settings, $k$ items are selected with the goal of representing the underlying distribution of claims, opinions, or characteristics in a large population. We study environments with two adversarial parties whose preferences over the selected items are commonly known and opposed. We propose the Quantile Mechanism: one party partitions the population into $k$ disjoint subsets, and the other selects one item from each subset. We show that this procedure is optimally representative among all feasible mechanisms, and illustrate its use in jury selection, multi-district litigation, and committee formation.

We consider the statistical problem of estimating constituent curves from observations of their aggregated curves, referred to as aggregated functional data, in models with strictly positive random errors following a Gamma distribution and correlated errors structured through AR(1) and ARFIMA processes. This problem arises in several areas of knowledge, such as chemometrics, for example, when absorbance curves of the constituents of a given substance must be estimated from its aggregated absorbance curve according to the Beer–Lambert law. In this context, we propose Bayesian wavelet-based methods to estimate the component functions within a functional data analysis framework. This approach has the advantage of accurately estimating curves with important local features, such as discontinuities, peaks, and oscillations, due to the representation properties of functions in wavelet bases. We further evaluate the performance of the proposed method through computational simulations, as well as applications to real data.

Estimating galaxy redshifts is crucial for constraining key physical quantities like those in the equation of state of dark energy. Modern telescopes such as the James Webb Space Telescope, the Euclid Space Telescope, and the NASA Nancy Grace Roman Space Telescope are producing massive amounts of spectroscopic data that enable precise redshift estimation. However, a galaxy's redshift can be estimated only when emission lines are present in the observed spectrum, which is unknown a priori. A novel Bayesian approach to estimating redshift and simultaneously testing for the presence of emission lines is developed. Although modern spectroscopic surveys involve millions of spectra and give rise to highly multimodal posterior distributions, the proposed framework remains computationally efficient, admitting a parallelizable implementation suitable for large-scale inference.

Discrete choice models are fundamental tools in management science, economics, and marketing for understanding and predicting decision-making. Logit-based models are dominant in applied work, largely due to their convenient closed-form expressions for choice probabilities. However, they impose restrictive assumptions on the stochastic utility component, constraining our ability to capture realistic substitution patterns. We propose an amortized inference approach that relies on a neural network emulator to approximate choice probabilities for general error distributions, including those with correlated errors. We develop a specialized neural network architecture designed to respect the invariance properties of discrete choice models. We provide group-theoretic foundations for the architecture, including a proof of universal approximation given a minimal set of invariant features. Once trained, the emulator enables rapid likelihood evaluation and gradient computation. We use Sobolev training, augmenting the likelihood loss with a gradient-matching penalty, so that the emulator learns both choice probabilities and their derivatives. We show that emulator-based maximum likelihood estimators are consistent and asymptotically normal under mild approximation conditions, and we provide sandwich standard errors that remain valid for a psuedo-true parameter even with imperfect likelihood approximation. Simulations show significant gains over the GHK simulator in accuracy and speed.

In deploying artificial intelligence (AI) models, selective prediction offers the option to abstain from making a prediction when uncertain about model quality. To fulfill its promise, it is crucial to enforce strict and precise error control over cases where the model is trusted. We propose Selective Conformal Risk control with E-values (SCoRE), a new framework for deriving such decisions for any trained model and any user-defined, bounded and continuously-valued risk. SCoRE offers two types of guarantees on the risk among “positive” cases in which the system opts to trust the model. Built upon conformal inference and hypothesis testing ideas, SCoRE first constructs a class of (generalized) e-values, which are non-negative random variables whose product with the unknown risk has expectation no greater than one. Such a property is ensured by data exchangeability without requiring any modeling assumptions. Passing these e-values on to hypothesis testing procedures, we yield the binary trust decisions with finite-sample error control. SCoRE avoids the need of uniform concentration, and can be readily extended to settings with distribution shifts. We evaluate the proposed methods with simulations and demonstrate their efficacy through applications to error management in drug discovery, health risk prediction, and large language models.

This article presents an argument for why quantum computers could unlock new methods for machine learning. We argue that spectral methods, in particular those that learn, regularise, or otherwise manipulate the Fourier spectrum of a machine learning model, are often natural for quantum computers. For example, if a generative machine learning model is represented by a quantum state, the Quantum Fourier Transform allows us to manipulate the Fourier spectrum of the state using the entire toolbox of quantum routines, an operation that is usually prohibitive for classical models. At the same time, spectral methods are surprisingly fundamental to machine learning: A spectral bias has recently been hypothesised to be the core principle behind the success of deep learning; support vector machines have been known for decades to regularise in Fourier space, and convolutional neural nets build filters in the Fourier space of images. Could, then, quantum computing open fundamentally different, much more direct and resource-efficient ways to design the spectral properties of a model? We discuss this potential in detail here, hoping to stimulate a direction in quantum machine learning research that puts the question of “why quantum?” first.

The autoresearch repository enables an LLM agent to optimize hyperparameters by editing training code directly. We use it as a testbed to compare classical HPO algorithms against LLM-based methods on tuning the hyperparameters of a small language model under a fixed compute budget. When defining a fixed search space over autoresearch, classical methods such as CMA-ES and TPE consistently outperform LLM-based agents, where avoiding out-of-memory failures matters more than search diversity. Allowing the LLM to directly edit source code narrows the gap to the classical methods but does not close it, even with frontier models available at the time of writing such as Claude Opus 4.6 and Gemini 3.1 Pro Preview. We observe that LLMs struggle to track optimization state across trials. In contrast, classical methods lack the domain knowledge of LLMs. To combine the strengths of both, we introduce Centaur, a hybrid that shares CMA-ES's interpretable internal state, including mean vector, step-size, and covariance matrix, with an LLM. Centaur achieves the best result in our experiments, and a 0.8B LLM already suffices to outperform all classical and pure LLM methods. Unconstrained code editing requires larger models to be competitive with classical methods. We further analyze search diversity, model scaling from 0.8B to frontier models, and ablate the fraction of LLM-proposed trials in Centaur. All in all, our results suggest that LLMs are most effective as a complement to classical optimizers, not as a replacement. Code is available at https://github.com/ferreirafabio/autoresearch-automl & interactive demo at https://ferreirafabio.github.io/autoresearch-automl.

Accurate inference on population dynamics, such as migration and changes in population size, is essential for policymaking, resource allocation and demographic research. Traditional censuses are expensive, infrequent and not timely, leading many countries to adopt register-based approaches to replace or complement them. A primary challenge is that such registers are incomplete: even when individuals are present, their activities may not generate records in specific registers, resulting in false negative observation error. Conversely, some registers arise from administrative or household-level processes, so that individuals may appear in registers despite being absent, leading to false positive observation error. Existing approaches often either rely on ad-hoc decisions that ignore one or both error types, offer inference on population snapshots but not dynamics, or are computationally too slow for practical use. We propose a scalable framework for inferring population size and dynamics from register data, building on Cormack-Jolly-Seber type capture-recapture models formulated as hidden Markov models. Inference is carried out using maximum likelihood estimation, with uncertainty quantified via the Bag of Little Bootstraps. The model accounts for temporary emigration, incorporates an arbitrary number of possibly interacting registers subject to both error types, and allows observation probabilities to vary with individual characteristics and unobservable heterogeneity. We illustrate the approach using Swedish population registers, where overcoverage - individuals registered as living in the country although they are no longer present - provides a motivating example. The application yields new insights into population dynamics and individual trajectories.

Consider two sequences of heterogeneous and independent portfolios of risks $T_1,T_2,\ldots$ and $T^*_{1}, T^*_{2},\ldots$ and, let $N_1$ and $N_2$ be two positive integer-valued random variables, independent of $T_i'$ and $T^*_i$, respectively. In this article, we investigate different stochastic inequalities involving $\min\{T_1,\ldots,T_{N_1}\}$ and $\min\{T^*_1,\ldots,T^*_{N_2}\},$ and $\max\{T_1,\ldots,T_{N_1}\}$ and $\max\{T^*_1,\ldots,T^*_{N_2}\}$ in the sense of usual stochastic order and reversed hazard rate order concerning maltivariate chain majorization order. These new results strengthen and generalize some of the well known results in the literature, including \cite{barmalzan2017ordering}, \cite{balakrishnan2018} and \cite{kundu2021_shock} for the case of random claim sizes. Different numerical examples are provided to highlight the applicability of this work. Finally, some interesting applications of our results in reliability theory and auction theory are presented.

The requirement of generating predictions that exactly fulfill the fundamental symmetry of the corresponding physical quantities has profoundly shaped the development of machine-learning models for physical simulations. In many cases, models are built using constrained mathematical forms that ensure that symmetries are enforced exactly. However, unconstrained models that do not obey rotational symmetries are often found to have competitive performance, and to be able to learn to a high level of accuracy an approximate equivariant behavior with a simple data augmentation strategy. In this paper, we introduce rigorous metrics to measure the symmetry content of the learned representations in such models, and assess the accuracy by which the outputs fulfill the equivariant condition. We apply these metrics to two unconstrained, transformer-based models operating on decorated point clouds (a graph neural network for atomistic simulations and a PointNet-style architecture for particle physics) to investigate how symmetry information is processed across architectural layers and is learned during training. Based on these insights, we establish a rigorous framework for diagnosing spectral failure modes in ML models. Enabled by this analysis, we demonstrate that one can achieve superior stability and accuracy by strategically injecting the minimum required inductive biases, preserving the high expressivity and scalability of unconstrained architectures while guaranteeing physical fidelity.