archive
Every paper Pith has read. Search by title, abstract, or pith.
883 papers in math.ST · page 15
-
Trajectory fitting recovers drift in singular Volterra equations
Drift estimation for rough processes under small noise asymptotic : trajectory fitting method
-
Diffusion models learn high-variance modes first via inverse scaling
An Analytical Theory of Spectral Bias in the Learning Dynamics of Diffusion Models
-
Path regularization bounds generalization error in finite deep nets
Path Regularization: A Near-Complete and Optimal Nonasymptotic Generalization Theory for Multilayer Neural Networks and Double Descent Phenomenon
-
Single-index model achieves 1D optimal rates for batched bandits
Batched Single-Index Global Multi-Armed Bandits with Covariates
-
Exact multi-graph alignment possible above critical threshold
The feasibility of multi-graph alignment: a Bayesian approach
-
Multi-graph alignment has sharp all-or-nothing threshold in Gaussian model
The feasibility of multi-graph alignment: a Bayesian approach
-
MinimaxSplit trees bound excess risk by worst-case child impurities
Stabilizing the Splits through Minimax Decision Trees
-
Bias-corrected top-two stats give consistent extreme estimates
Extreme Value Analysis based on Blockwise Top-Two Order Statistics
-
Low-degree polynomials fail below BBP and Kesten-Stigum thresholds
Sharp Phase Transitions in Estimation with Low-Degree Polynomials
-
Nyström MMD test matches minimax optimal separation rate
A Scalable Nystrom-Based Kernel Two-Sample Test with Permutations
-
LLMs can drive hallucination probability to near zero with better data
Hallucinations are inevitable but can be made statistically negligible
-
Dirichlet-kernel Gasser-Müller estimator introduced for simplex regression
On the Dirichlet-kernel Gasser--M\"uller estimator and its competitors for fixed design regression on the simplex
-
SGD multiplier bootstrap hits 1/sqrt(n) convex distance rate
Gaussian Approximation and Multiplier Bootstrap for Stochastic Gradient Descent
-
L1 quantile process convergence requires local absolute continuity
A necessary and sufficient condition for convergence in distribution of the quantile process in $L^1(0,1)$
-
Point cloud estimators for Minkowski tensors become asymptotically unbiased
Minkowski tensors for point clouds and voxelized data: robust, asymptotically unbiased estimators
-
Log-likelihood strongly concave near true tree parameters
Likelihood landscape of binary latent model on a tree
-
Heat split into two parts tracks free-energy gain in driven dot
Heat-dissipation decomposition and free-energy generation in a non-equilibrium dot with multi-electron states
-
Kernel estimators recover varifolds from point samples
A varifold-type estimation for data sampled on a rectifiable set
-
Minimum Hellinger distance estimators can be made private under a new Hellinger…
Private Minimum Hellinger Distance Estimation via Hellinger Distance Differential Privacy
-
Bayesian optimal transport partitions sphere for local regressions
Bayesian Sphere-on-Sphere Regression with Optimal Transport Maps
-
Unified method yields valid confidence sets for stochastic optimization
Honest Inference for Stochastic Optimization
-
Noisy SIS epidemic mixes in Theta(n log n) time
Mixing time for an epidemic model on graphs with external sources of infection
-
Pairwise classification learns ordered smooth boundaries with ReLU nets
Statistical learnability of smooth boundaries via pairwise binary classification with deep ReLU networks
-
Copula loss lifts Vision Transformer performance on both-eye myopia scans
Copula-enhanced Vision Transformer for high myopia diagnosis through OU UWF fundus images
-
Swapping test controls Type I error exactly for conditional independence
Testing conditional independence under isotonicity
-
Persistent homology yields stable time-series similarity score
A Stable Measure of Similarity for Time Series using Persistent Homology
-
ESPRIT DoA estimates become inconsistent for large arrays
A Large-Dimensional Analysis of ESPRIT DoA Estimation: Inconsistency and a Correction via RMT
-
A necessary and sufficient condition determines size controllability for…
A Necessary and Sufficient Condition for Size Controllability of Heteroskedasticity Robust Test Statistics
-
Rescaled entropic self-transport converges to score function
The entropic optimal (self-)transport problem: Limit distributions for decreasing regularization with application to score function estimation
-
Probit and logit DA samplers mix in linear time with sample size
Fast Mixing of Data Augmentation Algorithms: Bayesian Probit, Logit, and Lasso Regression
-
Known strata distributions still leave principal causal effects only partially identified
Partial identification of principal causal effects under violations of principal ignorability
-
DM embedding error decays as (log n/n) to 1/(8d+16)
How well behaved is finite dimensional Diffusion Maps?
-
Generalized ψ-estimators admit monotone decreasing form
Monotone representation and measurability of generalized $\psi$-estimators
-
Coin betting is optimal for e-variable mean testing
On the optimality of coin-betting for mean estimation
-
Communities recovered from binary activity when T ≫ N
Community detection for binary graphical models in high dimension
-
Book unifies proofs for conformal prediction guarantees
Theoretical Foundations of Conformal Prediction
-
Formula identifies causal effects from remote sensing by merging experiment and data
Program Evaluation with Remotely Sensed Outcomes
-
Wasserstein depth ranks distributions by centrality
Wasserstein Spatial Depth
-
Cross-fitting with out-of-sample losses fixes changepoint detection in complex models
Changepoint Detection in Complex Models: Cross-Fitting Is Needed
-
Matching estimators obey non-normalized CLT with explicit variance
Limit theorems of matching estimators with a fixed number of matches
-
Bayesian latent knockoffs control FDR at any level
Bayesian Controlled FDR Variable Selection via Parameter-Expanded Latent Knockoffs
-
Sp-GD recovers max-affine parameters from O(s log(d/s)) samples
Sparse Max-Affine Regression
-
kNN Laplacians reach O(N^{-2/(d+6)}) convergence on manifolds
Improved convergence rate of kNN graph Laplacians: differentiable self-tuned affinity
-
Layered clusters improve tail exponent estimates with missing data
Layered Hill estimator for extreme data in clusters
-
Batch variance update beats Ross method unless batch size is one
Extending Sheldon M. Ross's Method for Efficient Large-Scale Variance Computation
-
Undetectable corruption radius is twice Gaussian width for symmetric sets
Can we spot a fake?
-
Any conditional independence test yields a conditional two-sample test
General Frameworks for Conditional Two-Sample Testing
-
Non-central matrix ratio distribution derived
On non-central distribution of the matrix ratio
-
Minimax mean estimation error under arbitrary missingness decomposes into MCAR quantiles…
Estimation beyond Missing (Completely) at Random
-
Staleness factor model removes bias from high-frequency co-volatility estimates
Staleness Factors and Volatility Estimation at High Frequencies