Papers

The ISOKANN (Invariant Subspaces of Koopman Operators Learned by Artificial Neural Networks) framework provides a data-driven route to extract collective variables (CVs) and effective dynamics from complex molecular systems. In this work, we integrate the theoretical foundation of Koopman operators with Krylov-like subspace algorithms, and reduced dynamical modeling to build a coherent picture of how to describe metastable transitions in high-dimensional systems based on CVs. Starting from the identification of CVs based on dominant invariant subspaces, we derive the corresponding effective dynamics on the latent space and connect these to transition rates and times, committor functions, and transition pathways. The combination of Koopman-based learning and reduced-dimensional effective dynamics yields a principled framework for computing transition rates and pathways from simulation data. Numerical experiments on one-, two-, and three-dimensional benchmark potentials illustrate the ability of ISOKANN to reconstruct the coarse-grained kinetics and reproduce transition times across enthalpic and entropic barriers.

Reliability-based design optimization (RBDO) is traditionally formulated as a nested optimization and reliability problem and remains computationally demanding, especially in high dimensions. This paper proposes a new RBDO framework based on a stochastic simulator viewpoint, in which the deterministic limit-state function and uncertain model inputs are combined into a unified stochastic representation. For a given design, the system response is characterized directly through its conditional output distribution rather than through an explicit limit-state function. Stochastic emulators are constructed in the design space to approximate this conditional distribution, enabling semi-analytical evaluation of failure probabilities or associated quantiles without Monte Carlo simulation. Two approaches are considered: generalized lambda models (GLaM) and stochastic polynomial chaos expansions (SPCE). Both yield deterministic mappings between design variables and reliability constraints, thereby eliminating the classical double-loop structure and enabling standard deterministic optimization. The approach is assessed on benchmark problems ranging from low to very high dimension, including stochastic excitation, and compared with Kriging in the full input space and heteroscedastic Gaussian processes in the design space. The proposed method provides substantial computational gains, particularly for high-dimensional random inputs, while retaining comparable efficiency to Kriging in low dimensions. Unlike heteroscedastic Gaussian processes, GLaM and SPCE also avoid restrictive assumptions on the conditional response distribution.

There is a growing demand for efficient data removal to comply with regulations like the GDPR and to mitigate the influence of biased or corrupted data. This has motivated the field of machine unlearning, which aims to eliminate the influence of specific data subsets without the cost of full retraining. In this work, we propose a statistical framework for machine unlearning with generic loss functions and establish theoretical guarantees. For squared loss, especially, we develop Unlearning Least Squares (ULS) and establish its minimax optimality for estimating the model parameter of remaining data when only the pre-trained estimator, forget samples, and a small subsample of the remaining data are available. Our results reveal that the estimation error decomposes into an oracle term and an unlearning cost determined by the forget proportion and the forget model bias. We further establish asymptotically valid inference procedures without requiring full retraining. Numerical experiments and real-data applications demonstrate that the proposed method achieves performance close to retraining while requiring substantially less data access.

We develop methods for estimating how infinitesimal policy changes affect long-term outcomes in dynamic systems. We show that dynamic marginal policy effects (MPEs) can be identified via tractable reduced-form expressions, and can be estimated under a general sequential unconfoundedness assumption. We also propose a doubly robust estimator for dynamic MPEs. Our approach does not require observing full dynamic state information (as is typically assumed for off-policy evaluation in Markov decision processes), and does not incur an exponential curse of horizon (as is typical in non-Markovian off-policy evaluation). We demonstrate practicality and robustness of our approach in a number of simulations, including one motivated by a dynamic pricing application where people use past prices to form a reference level for current prices.

We propose an active learning algorithm for linear system identification with optimal centered noise excitation. Notably, our algorithm, based on ordinary least squares and semidefinite programming, attains the minimal sample complexity while allowing for efficient computation of an estimate of a system matrix. More specifically, we first establish lower bounds of the sample complexity for any active learning algorithm to attain the prescribed accuracy and confidence levels. Next, we derive a sample complexity upper bound of the proposed algorithm, which matches the lower bound for any algorithm up to universal factors. Our tight bounds are easy to interpret and explicitly show their dependence on the system parameters such as the state dimension.

Clustering is viewed as an unsupervised technique, but in practice it requires guidance to uncover meaningful structures. We formalize this with guided clustering, a paradigm that uses a guiding variable to steer the discovery process, and introduce the Guided Clustering Variational Autoencoder (GCVAE) as its deep generative realization. GCVAE learns a latent space structured as a Gaussian Mixture Model by optimizing a variational objective that forces the representation to be maximally informative about the guiding variable. This framework allows the resulting clustering to be reoriented by changing the guiding variable, yielding clusters that are meaningful for the specified context. Experiments on public (MNIST-SVHN) and proprietary connected health devices data demonstrate GCVAE's ability to discover coherent and task-relevant clusters in complex settings.

We consider the problem of evaluating black-box multi-class classifiers. In the standard setup, we observe class labels $Y\in \{0,1,\ldots,M-1\}$ generated according to the conditional distribution $ Y|X \sim \text{ Multinom}\big(\eta(X)\big), $ where $X$ denotes the features and $\eta$ maps from the feature space to the $(M-1)$-dimensional simplex. A black-box classifier is an estimate $\hat{\eta}$ for which we make no assumptions about the training algorithm. Given holdout data, our goal is to evaluate the performance of the classifier $\hat{\eta}$. Recent work suggests treating this as a goodness-of-fit problem by testing the hypothesis $H_0: \rho((X,Y),(X',Y')) \le \delta$, where $\rho$ is some metric between two distributions, and $(X',Y')\sim P_X\times \text{ Multinom}(\hat\eta(X))$. Combining ideas from algorithmic fairness, Neyman-Pearson lemma, and conformal p-values, we propose a new methodology for this testing problem. The key idea is to generate a second sample $(X',Y') \sim P_X \times \text{ Multinom}\big(\hat\eta(X)\big)$ allowing us to reduce the task to two-sample conditional distribution testing. Using part of the data, we train an auxiliary binary classifier called a distinguisher to attempt to distinguish between the two samples. The distinguisher's ability to differentiate samples, measured using a rank-sum statistic, is then used to assess the difference between $\hat{\eta}$ and $\eta$ . Using techniques from cross-validation central limit theorems, we derive an asymptotically rigorous test under suitable stability conditions of the distinguisher.

We study language models as evolving model organisms and ask when autoregressive next-token learning selects for world-tracking representations. For any encoding of latent world states, the Bayes-optimal next-token cross-entropy decomposes into the irreducible conditional entropy plus a Jensen–Shannon excess term. That excess vanishes if and only if the encoding preserves the training ecology's equivalence classes. This yields a precise notion of ecological veridicality for language models and identifies the minimum-complexity zero-excess solution as the quotient partition by training equivalence. We then determine when this fixed-encoding analysis applies to transformer families: frozen dense and frozen Mixture-of-Experts transformers satisfy it, in-context learning does not enlarge the model's separation set, and per-task adaptation breaks the premise. The framework predicts two characteristic failure modes: simplicity pressure preferentially removes low-gain distinctions, and training-optimal models can still incur positive excess on deployment ecologies that refine the training ecology. A conditional dynamic extension shows how inter-model selection and post-training can recover such gap distinctions under explicit heredity, variation, and selection assumptions. Exact finite-ecology checks and controlled microgpt experiments validate the static decomposition, split-merge threshold, off-ecology failure pattern, and two-ecology rescue mechanism in a regime where the relevant quantities are directly observable. The goal is not to model frontier systems at scale, but to use small language models as laboratory organisms for theory about representational selection.

Multimodal representation learning is commonly built on a shared-private decomposition, treating latent information as either common to all modalities or specific to one. This binary view is often inadequate: many factors are shared by only subsets of modalities, and ignoring such partial sharing can over-align unrelated signals and obscure complementary information. We propose Hierarchical Contrastive Learning (HCL), a framework that learns globally shared, partially shared, and modality-specific representations within a unified model. HCL combines a hierarchical latent-variable formulation with structural sparsity and a structure-aware contrastive objective that aligns only modalities that genuinely share a latent factor. Under uncorrelated latent variables, we prove identifiability of the hierarchical decomposition, establish recovery guarantees for the loading matrices, and derive parameter estimation and excess-risk bounds for downstream prediction. Simulations show accurate recovery of hierarchical structure and effective selection of task-relevant components. On multimodal electronic health records, HCL yields more informative representations and consistently improves predictive performance.

Large language model (LLM) evaluation platforms increasingly rely on pairwise human judgments. These data are noisy, sparse, and non-uniform, yet leaderboards are reported with limited uncertainty quantification. We study this as semiparametric inference for a low-rank latent score tensor observed through pairwise comparisons under Bradley-Terry-Luce-type models. This places LLM evaluation in a new tensor completion setting with structured observations, non-uniform sampling, and pairwise contrasts. Our target is a smooth functional $\psi(T^\star)$, including linear estimands such as ability gaps and nonlinear ones such as win probabilities. We derive the information operator on the low-rank tangent space, the efficient influence function, and the semiparametric efficiency bound, then construct a one-step debiased estimator with asymptotic normality. A central challenge is that the information operator is anisotropic and does not commute with the tangent-space projection, creating a bottleneck absent from isotropic models. We introduce a score-whitening method that equalizes local Fisher information and restores stable inference at the optimal sample-complexity scale. Our results provide a principled framework for uncertainty quantification in LLM evaluation and more broadly for inference on low-rank structures from pairwise data.

Obtaining high-quality labels is costly, whereas unlabeled covariates are often abundant, motivating semi-supervised inference methods with reliable uncertainty quantification. Prediction-powered inference (PPI) leverages a machine-learning predictor trained on a small labeled sample to improve efficiency, but it can lose efficiency under model misspecification and suffer from coverage distortions due to label reuse. We introduce Machine-Learning-Assisted Generalized Entropy Calibration (MEC), a cross-fitted, calibration-weighted variant of PPI. MEC improves efficiency by reweighting labeled samples to better align with the target population, using a principled calibration framework based on Bregman projections. This yields robustness to affine transformations of the predictor and relaxes requirements for validity by replacing conditions on raw prediction error with weaker projection-error conditions. As a result, MEC attains the semiparametric efficiency bound under weaker assumptions than existing PPI variants. Across simulations and a real-data application, MEC achieves near-nominal coverage and tighter confidence intervals than CF-PPI and vanilla PPI.

The classical Gaussian mixture model assumes homogeneity within clusters, an assumption that often fails in real-world data where observations naturally exhibit varying scales or intensities. To address this, we introduce the individual-heterogeneous sub-Gaussian mixture model, a flexible framework that assigns each observation its own heterogeneity parameter, thereby explicitly capturing the heterogeneity inherent in practical applications. Built upon this model, we propose an efficient spectral method that provably achieves exact recovery of the true cluster labels under mild separation conditions, even in high-dimensional settings where the number of features far exceeds the number of samples. Numerical experiments on both synthetic and real data demonstrate that our method consistently outperforms existing clustering algorithms, including those designed for classical Gaussian mixture models.

Sampling physical systems with rough energy landscapes is hindered by rare events and metastable trapping. While Boltzmann generators already offer a solution, their reliance on the reverse Kullback–Leibler divergence frequently induces catastrophic mode collapse, missing specific modes in multi-modal distributions. Here, we introduce the Jeffreys Flow, a robust generative framework that mitigates this failure by distilling empirical sampling data from Parallel Tempering trajectories using the symmetric Jeffreys divergence. This formulation effectively balances local target-seeking precision with global modes coverage. We show that minimizing Jeffreys divergence suppresses mode collapse and structurally corrects inherent inaccuracies via distillation of the empirical reference data. We demonstrate the framework's scalability and accuracy on highly non-convex multidimensional benchmarks, including the systematic correction of stochastic gradient biases in Replica Exchange Stochastic Gradient Langevin Dynamics and the massive acceleration of exact importance sampling in Path Integral Monte Carlo for quantum thermal states.

Ordinary differential equations (ODEs) provide a powerful framework for modeling dynamic systems arising in a wide range of scientific domains. However, most existing ODE methods focus on a single system, and do not adequately address the problem of learning shared patterns from multiple heterogeneous dynamic systems. In this article, we propose a novel distributionally robust learning approach for modeling heterogeneous ODE systems. Specifically, we construct a robust dynamic system by maximizing a worst-case reward over an uncertainty class formed by convex combinations of the derivatives of trajectories. We show the resulting estimator admits an explicit weighted average representation, where the weights are obtained from a quadratic optimization that balances information across multiple data sources. We further develop a bi-level stabilization procedure to address potential instability in estimation. We establish rigorous theoretical guarantees for the proposed method, including consistency of the stabilized weights, error bound for robust trajectory estimation, and asymptotical validity of pointwise confidence interval. We demonstrate that the proposed method considerably improves the generalization performance compared to the alternative solutions through both extensive simulations and the analysis of an intracranial electroencephalogram data.

The sufficient cause framework has been used for decades to improve our understanding of both basic and more complex causal concepts in epidemiology, such as mediation and interaction. Here, we make use of this framework to provide a description of truncation by death, in which the outcome of interest is undefined for individuals who die before the time of assessment at the end of follow-up. We explain the non-causal nature of the crude estimand that compares outcomes by treatment levels conditional on observed survival by showing that it corresponds to a comparison of distinct risk status types, which are defined based on the susceptibility to sufficient causes. Further, expressions for the crude estimand and for the survivor average causal effect, a causal estimand defined under the principal stratification approach, are provided in terms of population-level joint frequencies of the background factors of sufficient causes. Finally, we also describe conditions, based on background factors of sufficient causes, under which the survivor average causal effect is null. Our description of this problem, which studies truncation by death from a new perspective, might encourage further analyses of principal stratification-based estimands using sufficient causes.

A statistical analysis of precipitation at Rio Grande do Sul State was presented in this article. The aim of this work was to identify spatial and temporal patterns of maximum precipitation, which was achieved by fitting a theoretical variogram in maximum annual rainfalls and their times of occurrence. In the literature, it was found that this pattern occurs according to phenomena typical from middle latitude, such as low and high level jets, and interactions between them. Some years ago, the relationship between maximum annual rainfalls and synoptic predominant configurations was found. Therefore, this work sought to understand the climatic characteristics that are important in Aerospace and Aeronautics, as extreme weather can cause numerous consequences in these activities. The use and validation of this proposed method would make possible its application in other regions of interest in the Brazilian aerospace. Understanding these climatological features of the atmospheric circulation dynamics, and analyzing maximum annual rainfall would allow a more efficient and appropriate climate trend forecast and its application in aerospace activities.

Sparse Canonical Correlation Analysis (SCCA) is a fundamental statistical tool for identifying linear relationships in high-dimensional, multi-view data. While minimax theory establishes an optimal sample complexity scaling additively with the sparsity levels of the canonical vectors, computationally efficient algorithms typically suffer from a suboptimal multiplicative dependence. This computational-statistical gap is intrinsically tied to worst-case “flat” signal assumptions. In practice, however, multi-view signals frequently exhibit structured energy concentration, such as a power-law decay. To exploit this structural concentration and bypass the worst-case bottleneck, we propose Bilateral Spectral Energy Pursuit (Bi-SEP). Operating directly on the cross-covariance matrix, Bi-SEP is a stagewise adaptive algorithm that utilizes a proxy refinement step to dynamically track and capture cross-view signal energy. Theoretically, we establish a profile-adaptive sample complexity bound governed by the coupled energy profiles of the two views. Notably, under power-law decay models, we reveal a synergistic phase transition: the optimal linear sample complexity is attainable provided that the aggregate decay rate of the two views is sufficiently large. This result demonstrates that a highly concentrated signal in one view allows the model to accommodate a completely flat signal in its partner. Numerical experiments validate our theoretical findings, illustrating the advantages of Bi-SEP in structured, non-flat signal regimes.

Time series foundation models (TSFMs) are widely used as generic feature extractors, yet the notion of non-stationarity in their embedding spaces remains poorly understood. Recent work often conflates non-stationarity with distribution shift, blurring distinctions fundamental to classical time-series analysis and long-standing methodologies such as statistical process control (SPC). In SPC, non-stationarity signals a process leaving a stable regime - via shifts in mean, variance, or emerging trends - and detecting such departures is central to quality monitoring and change-point analysis. Motivated by this diagnostic tradition, we study how different forms of distributional non-stationarity - mean shifts, variance changes, and linear trends - become linearly accessible in TSFM embedding spaces under controlled conditions. We further examine temporal non-stationarity arising from persistence, which reflects violations of weak stationarity due to long-memory or near-unit-root behavior rather than explicit distributional shifts. By sweeping shift strength and probing multiple TSFMs, we find that embedding-space detectability of non-stationarity degrades smoothly and that different models exhibit distinct, model-specific failure modes.

This article presents the results of a data science study conducted at a container terminal, aimed at reducing unproductive container moves through the prediction of service requirements and container dwell times. We develop and evaluate machine learning models that leverage historical operational data to anticipate which containers will require pre-clearance handling services prior to cargo release and to estimate how long they are expected to remain in the terminal. As part of the data preparation process, we implement a classification system for cargo descriptions and perform deduplication of consignee records to improve data consistency and feature quality. These predictive capabilities provide valuable inputs for strategic planning and resource allocation in yard operations. Across multiple temporal validation periods, the proposed models consistently outperform existing rule-based heuristics and random baselines in precision and recall. These results demonstrate the practical value of predictive analytics for improving operational efficiency and supporting data-driven decision-making in container terminal logistics.

Strict minimum message length (SMML) is an information-theoretic coding principle that represents a continuous statistical model by a finite set of assertions and a partition of the sample space. We show that the SMML objective decomposes into assertion entropy and conditional cross-entropy, balancing the cost of identifying an assertion against the cost of encoding data under the assigned model. For any fixed partition, the optimal codepoint for each cell is the model distribution that minimises Kullback-Leibler divergence from the data distribution restricted to that cell. Using the local Fisher-Rao geometry of regular parametric models, we show that, under high-resolution regularity conditions, optimal SMML partitions are asymptotically the pullback, through the maximum likelihood estimator, of weighted Fisher-Rao Voronoi tessellations in parameter space, with assertion probabilities appearing as additive weights. For regular exponential families, SMML codepoints satisfy a moment-matching condition and admit an interpretation as KL/Bregman centroids, while exact SMML cells are pullbacks of convex polyhedra in sufficient-statistic space. Together, these results show that SMML induces a natural information-geometric quantisation linking entropy-based coding, KL projection, and divergence-based Voronoi geometry.

Preprocessing leakage arises when scaling, imputation, or other data-dependent transformations are estimated before resampling, inflating apparent performance while remaining hard to detect. We present fastml, an R package that provides a single-call interface for leakage-aware machine learning through guarded resampling, where preprocessing is re-estimated inside each resample and applied to the corresponding assessment data. The package supports grouped and time-ordered resampling, blocks high-risk configurations, audits recipes for external dependencies, and includes sandboxed execution and integrated model explanation. We evaluate fastml with a Monte Carlo simulation contrasting global and fold-local normalization, a usability comparison with tidymodels under matched specifications, and survival benchmarks across datasets of different sizes. The simulation demonstrates that global preprocessing substantially inflates apparent performance relative to guarded resampling. fastml matched held-out performance obtained with tidymodels while reducing workflow orchestration, and it supported consistent benchmarking of multiple survival model classes through a unified interface.

Complexity researchers view burstiness–fluctuating levels of activity–as evidence of hidden interactions within the system generating the activity signal. Yet, current burstiness metrics miss evidence of burstiness in some moderately bursty distributions and under moderate sampling conditions. The canonical Burstiness Parameter (BP) compares distributions of timing statistics to the exponential distribution, representing the timing of independent random events, but it provides false negatives for some parameter ranges of power laws, with and without cut-offs. We introduce a metric that maintains BP's measurement approach but reduces false negatives: the Burstiness Tail-based Index (BTI). Based on ratios of differences in quantiles, BTI correctly classifies bursty distributions over certain parameter ranges misclassified by BP. Additionally, we find BTI to be more robust than BP in the presence of limited sample sizes and short observation windows, using simulated samples drawn from distributions correctly classified by BP in their analytical form. As a case study, we revisit an analysis of human activity data and find that the choice of BTI over BP influences interpretations of the timescales of burstiness in the dataset. Given these analytical, simulated, and empirical results, we argue for BTI's practical advantage over BP in assessing burstiness in real-world temporal signals for complexity research and time series modeling.

We study the problem of robust mean estimation with adversarially contaminated data under star-shaped constraints in a heavy-tailed noise setting, where only a finite second moment $ \sigma ^2 $ is assumed. For a contamination level $ \varepsilon$ below some constant, we show that the minimax rate of the squared $ \ell_2 $ loss is $ \max( \delta ^{*2}, \varepsilon \sigma ^2) \wedge d^2 $ for a star-shaped set with diameter $ d $ (set $d = \infty$ if the set is unbounded), with $ \delta ^* $ determined via the local entropy $ \log M^\mathrm{ loc }(\delta ,c) $ as \begin{align*} \delta ^*:= \sup\bigg\{\delta \geq 0: N\frac{\delta ^2}{\sigma ^2}\leq \log M^\mathrm{ loc }(\delta ,c) \bigg\}, \end{align*} where $ c $ is a sufficiently large constant. Crucially, we require that the sample size satisfies $N \gtrsim \mathop{ \sup }\limits_{\delta \geq 0} \log M^\mathrm{ loc }(\delta ,c)$. We also show that the minimax rate is $ \max(\delta^{*2},\varepsilon ^2\sigma ^2) \wedge d^2 $ for known or sign-symmetric distributions, matching the rate achieved in the Gaussian case.

Blind-spot mass is a Good-Turing framework for quantifying deployment coverage risk in machine learning. In modern ML systems, operational state distributions are often heavy-tailed, implying that a long tail of valid but rare states is structurally under-supported in finite training and evaluation data. This creates a form of 'coverage blindness': models can appear accurate on standard test sets yet remain unreliable across large regions of the deployment state space. We propose blind-spot mass B_n(tau), a deployment metric estimating the total probability mass assigned to states whose empirical support falls below a threshold tau. B_n(tau) is computed using Good-Turing unseen-species estimation and yields a principled estimate of how much of the operational distribution lies in reliability-critical, under-supported regimes. We further derive a coverage-imposed accuracy ceiling, decomposing overall performance into supported and blind components and separating capacity limits from data limits. We validate the framework in wearable human activity recognition (HAR) using wrist-worn inertial data. We then replicate the same analysis in the MIMIC-IV hospital database with 275 admissions, where the blind-spot mass curve converges to the same 95% at tau = 5 across clinical state abstractions. This replication across structurally independent domains - differing in modality, feature space, label space, and application - shows that blind-spot mass is a general ML methodology for quantifying combinatorial coverage risk, not an application-specific artifact. Blind-spot decomposition identifies which activities or clinical regimes dominate risk, providing actionable guidance for industrial practitioners on targeted data collection, normalization/renormalization, and physics- or domain-informed constraints for safer deployment.

We study hypothesis testing for penalized estimators in settings where the full marginal distribution of a multivariate response is difficult to specify, such as longitudinal data with correlated measurements or high-dimensional heteroscedastic regression. Assuming that the conditional mean model is correctly specified, we establish that the penalized estimating equations admit a $\sqrt{n}$-consistent solution, even when the working covariance structure is misspecified. Our inferential target is a low-dimensional subvector of parameters associated with the mean model. We show that the resulting test statistic converges to a $\chi^2$ distribution, and that its asymptotic power depends on the nuisance covariance function. To mitigate this dependence, we propose estimating the covariance function via cross-fitting, which provides a calibrated and robust procedure for inference.

Multi-channel sensor networks in industrial IoT often exceed available bandwidth. We propose PCA-Triage, a streaming algorithm that converts incremental PCA loadings into proportional per-channel sampling rates under a bandwidth budget. PCA-Triage runs in O(wdk) time with zero trainable parameters (0.67 ms per decision). We evaluate on 7 benchmarks (8–82 channels) against 9 baselines. PCA-Triage is the best unsupervised method on 3 of 6 datasets at 50% bandwidth, winning 5 of 6 against every baseline with large effect sizes (r = 0.71–0.91). On TEP, it achieves F1 = 0.961 +/- 0.001 – within 0.1% of full-data performance – while maintaining F1 > 0.90 at 30% budget. Targeted extensions push F1 to 0.970. The algorithm is robust to packet loss and sensor noise (3.7–4.8% degradation under combined worst-case).

This paper introduces a novel generative framework for synthesising forward-looking, càdlàg stochastic trajectories that are sequentially consistent with time-evolving path-law proxies, thereby incorporating anticipated structural breaks, regime shifts, and non-autonomous dynamics. By framing path synthesis as a sequential matching problem on restricted Skorokhod manifolds, we develop the Anticipatory Neural Jump-Diffusion (ANJD) flow, a generative mechanism that effectively inverts the time-extended Marcus-sense signature. Central to this approach is the Anticipatory Variance-Normalised Signature Geometry (AVNSG), a time-evolving precision operator that performs dynamic spectral whitening on the signature manifold to ensure contractivity during volatile regime shifts and discrete aleatoric shocks. We provide a rigorous theoretical analysis demonstrating that the joint generative flow constitutes an infinitesimal steepest descent direction for the Maximum Mean Discrepancy functional relative to a moving target proxy. Furthermore, we establish statistical generalisation bounds within the restricted path-space and analyse the Rademacher complexity of the whitened signature functionals to characterise the expressive power of the model under heavy-tailed innovations. The framework is implemented via a scalable numerical scheme involving Nyström-compressed score-matching and an anticipatory hybrid Euler-Maruyama-Marcus integration scheme. Our results demonstrate that the proposed method captures the non-commutative moments and high-order stochastic texture of complex, discontinuous path-laws with high computational efficiency.

Gradient normalization can stabilize deep-learning optimization, and spectral normalizations are especially natural for matrix-shaped parameter blocks; Muon is the motivating example. We study an idealized deterministic, continuous-time, vanishing-momentum version of this idea in the mean-field regime, where wide models are represented by probability measures on parameter space. Starting from normalized matrix flows, we introduce Spectral Wasserstein distances indexed by norms $\gamma$ on positive semidefinite matrices: the trace norm recovers the classical $W_2$ distance, the operator norm gives the Muon geometry, and Schatten norms interpolate between them. We develop a static Kantorovich formulation, a max–min robust-cost representation, and Gaussian reductions extending the Bures formula. For monotone norms, we prove equivalence with a Benamou–Brenier formulation and obtain a gradient-flow interpretation of normalized mean-field training dynamics. Numerical experiments illustrate the resulting flows for MMD objectives, Gaussian reductions, two-layer ReLU models, and shallow attention.

Tabular foundation models (TFMs) such as TabPFN (Tabular Prior-Data Fitted Network) are designed to generalize across heterogeneous tabular datasets through in-context learning (ICL). They perform prediction in a single forward pass conditioned on labeled examples without dataset-specific parameter updates. This paradigm is particularly attractive in industrial domains (e.g., finance and healthcare) where tabular prediction is pervasive. Retraining a bespoke model for each new table can be costly or infeasible in these settings, while data quality issues such as irrelevant predictors, correlated feature groups, and label noise are common. In this paper, we provide strong empirical evidence that TabPFN is highly robust under these sub-optimal conditions. We study TabPFN and its attention mechanisms for binary classification problems with controlled synthetic perturbations that vary: (i) dataset width by injecting random uncorrelated features and by introducing nonlinearly correlated features, (ii) dataset size by increasing the number of training rows, and (iii) label quality by increasing the fraction of mislabeled targets. Beyond predictive performance, we analyze internal signals including attention concentration and attention-based feature ranking metrics. Across these parametric tests, TabPFN is remarkably resilient: ROC-AUC remains high, attention stays structured and sharp, and informative features are highly ranked by attention-based metrics. Qualitative visualizations with attention heatmaps, feature-token embeddings, and SHAP plots further support a consistent pattern across layers in which TabPFN increasingly concentrates on useful features while separating their signals from noise. Together, these findings suggest that TabPFN is a robust TFM capable of maintaining both predictive performance and coherent internal behavior under various scenarios of data imperfections.

Sparse identification of nonlinear dynamics (SINDy) has been widely used to discover the governing equations of a dynamical system from data. It uses sparse regression techniques to identify parsimonious models of unknown systems from a library of candidate functions. Therefore, it relies on the assumption that the dynamics are sparsely represented in the coordinate system used. To address this limitation, one seeks a coordinate transformation that provides reduced coordinates capable of reconstructing the original system. Recently, SINDy autoencoders have extended this idea by combining sparse model discovery with autoencoder architectures to learn simplified latent coordinates together with parsimonious governing equations. A central challenge in this framework is robustness to measurement error. Inspired by noise-separating neural network structures, we incorporate a noise-separation module into the SINDy autoencoder architecture, thereby improving robustness and enabling more reliable identification of noisy dynamical systems. Numerical experiments on the Lorenz system show that the proposed method recovers interpretable latent dynamics and accurately estimates the measurement noise from noisy observations.