archive
Every paper Pith has read. Search by title, abstract, or pith.
883 papers in math.ST · page 14
-
Selective p-values fix inference after F-screening
Valid F-screening in linear regression
-
Compressed models learn sample-efficiently and generalize better
Efficient compression of neural networks and datasets
-
Liouville PDE removes diffusion from sliced Wasserstein flow
Liouville PDE-based sliced-Wasserstein flow
-
Reversed MDP reformulation solves policy testing with optimal samples
Policy Testing in Markov Decision Processes
-
Union of subspaces lets transformers generalize ICL to all angles
Out-of-Distribution Generalization of In-Context Learning: A Low-Dimensional Subspace Perspective
-
Interacting reinforced processes model patent category spillovers
Modeling Innovation Ecosystem Dynamics through Interacting Reinforced Bernoulli Processes
-
Residual encoder blocks negative transfer
Residual Feature Integration is Sufficient to Prevent Negative Transfer
-
Two-stage adaptive experiments get valid tests under weaker assumptions
Assumption-lean weak limits and tests for two-stage adaptive experiments
-
Neural nets achieve arbitrarily fast rates under hard margin
Super-fast Rates of Convergence for Neural Network Classifiers under the Hard Margin Condition
-
Gaussian approximations justify bootstrap for local SGD
Sharp Gaussian approximations for Decentralized Federated Learning
-
Checkerboard copulas bound Chatterjee's ξ from below with convergence
Measures of association for approximating copulas
-
Deepest scatter matrices yield explicit bias and breakdown curves
Bias robustness of depth estimators in multivariate settings
-
The paper introduces a semiparametric framework called D2S3 for semi-supervised…
Semiparametric semi-supervised learning for general targets under distribution shift and decaying overlap
-
Bounds derived for approximating number of sample maxima
Approximations for the number of maxima and near-maxima in independent data
-
Estimators trace treatment effects along assignment boundaries
Estimation and Inference in Boundary Discontinuity Designs: Location-Based Methods
-
FHTD selects unit-root models consistently with many predictors
Model Selection for Unit-root Time Series with Many Predictors
-
Kernel embeddings turn measure equality into Gaussian singularity test
Kernel Embeddings and the Separation of Measure Phenomenon
-
Transformers bound errors by task manifold dimension
Transformers for Learning on Noisy and Task-Level Manifolds: Approximation and Generalization Insights
-
Log-factor overparametrization unlocks optimal recovery in phase retrieval
Phase retrieval and matrix sensing via benign and overparametrized nonconvex optimization
-
Closed forms specify pilot accuracy for marginal optimization
Marginal minimization and sup-norm expansions in perturbed optimization
-
Expected weighted designs protect mixed-factor experiments from parameter errors
Expected Weighted D-optimal Designs for Experiments with Mixed Factors
-
Hypergeometric forms give root distributions for group mean ratios
On incomplete Gamma and Beta integrals
-
One realization now suffices for conditional independence testing in nonlinear nonstation
Conditional independence testing with a single realization of a multivariate nonstationary nonlinear time series
-
Bahadur efficiency lower bound extends to moderate deviations
Bahadur asymptotic efficiency in the zone of moderate deviation probabilities
-
Regularized SGD learns operators with dimension-free rates
Learning Operators by Regularized Stochastic Gradient Descent with Operator-valued Kernels
-
Scalar gauges on support functions yield set-valued conditionals for random sets
Set-valued conditional functionals of random sets
-
Entropy conditions ensure intersection and composition for discrete variables
On the Intersection and Composition properties of conditional independence
-
Proxies identify indirect effects under full confounding
Proximal Inference for Indirect and Intervening Effects in Population Interventions
-
Any VC class is learnable from positive samples under smoothness
Smoothed Analysis of Learning from Positive Samples
-
Theory merger yields guarantees for learning from rare tail data
Weak Signals and Heavy Tails: Learning Theory meets Extreme Value Analysis
-
Tangent method extends variational inference to super-Gaussian models
A Generalized Tangent Approximation based Variational Inference Framework for Strongly Super-Gaussian Likelihoods
-
Marchenko-Pastur inversion achieves O(n^{-1+ε}) rate for spectral stats
Estimation of Population Linear Spectral Statistics by Marchenko--Pastur Inversion
-
Risk formulas derived for ridge regression under varying variances
High-dimensional ridge regression with random features for non-identically distributed data with a variance profile
-
ROC and PR curves collapse to one function G of score CDFs
On the Geometry of Receiver Operating Characteristic and Precision-Recall Curves
-
Proper scoring rules enable consistent estimation and forecast evaluation
Proper scoring rules for estimation and forecast evaluation
-
Projected forward model yields optimal low-rank posterior
Optimal low-rank posterior mean and distribution approximation in linear Gaussian inverse problems on Hilbert spaces
-
Finite-difference scores rank variables stably under noise
Global Activity Scores
-
Axioms for empirical sampling yield abstract Glivenko-Cantelli theorem
Empirical Measures and Strong Laws of Large Numbers in Categorical Probability
-
KL divergence DRO optimally estimates heavy-tailed means
Robust Mean Estimation for Optimization: The Impact of Heavy Tails
-
Sliced Wasserstein distance obeys CLT centered at population value
An improved central limit theorem for the empirical sliced Wasserstein distance
-
Sellers converge to Nash prices at O(T^{-1/7}) with O(T^{5/7}) regret
Revenue Maximization Under Sequential Price Competition Via The Estimation Of s-Concave Demand Functions
-
Lower bound sets minimum repeats for reliable optimizer metrics
A Statistical Analysis for Per-Instance Evaluation of Stochastic Optimizers: Avoiding Unreliable Conclusions
-
MLE for distributional regression stays consistent under random censoring
Asymptotic properties of the MLE in distributional regression under random censoring
-
Fixed projections recover multivariate mixture parameters
Two statistical problems for multivariate mixture distributions
-
Chatterjee xi continuous under Markov products
On continuity of Chatterjee's rank correlation and related dependence measures
-
Condorcet cycles block reward-based LLM alignment almost surely
Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibrium
-
Low-rank graphon estimates control welfare loss in games
Low-Rank Graphon Estimation: Theory and Applications to Graphon Games
-
CUSUM detectors hit optimal rates for jumps and kinks online
Online jump and kink detection in segmented linear regression: Statistical optimality meets computational efficiency
-
Adaptive sampling guarantees error bound for relative risk
Estimation of relative risk, odds ratio and their logarithms with guaranteed accuracy and controlled sample size ratio
-
Early-stopped mirror descent matches sharp LSE risk bounds
Sharp Risk Bounds for Early-Stopping in Gaussian Linear Regression