arXiv subjects

Xiaohan Hu

Publications and source records attributed to Xiaohan Hu.

6 recordsLinked to original sources

arXiv2026

Signal-to-Noise Ratio Inference for Multivariate High-Dimensional Linear Models

We study inference for the fraction of total response variation explained by multivariate high-dimensional linear models. Two moment equations yield explicit estimates without fitting the regression coefficients. We establish asymptotically valid confidence intervals for fixed- and random-effects models, allowing the response dimension to grow with the sample size. The theory describes how response dependence affects precision and provides a correction for unequal noise levels in the random-design setting. Simulations assess finite-sample performance, and an analysis of yeast growth across environments illustrates joint inference for additive marker-based signal.

math.ST↗

arXiv2026

Misspecification Analysis of High-Dimensional Random Effects Models for Estimation of Signal-to-Noise Ratios

Estimation of signal-to-noise ratios and residual variances in high-dimensional linear models has various important applications, including heritability estimation in bioinformatics. One widely used estimator is the Gaussian random-effects maximum likelihood estimator (MLE), based on the likelihood of the homogeneous Gaussian random-effects model in which both the regression coefficients and the noise variables are assumed to be i.i.d. Gaussian. This paper studies the behavior of this likelihood estimator under model misspecification. For isotropic random designs with independent, symmetric, sub-Gaussian entries, we establish consistency and asymptotic normality of the SNR MLE for fixed dense coefficient vectors and independent, centered, heteroscedastic finite-moment noise, allowing moderately heavy-tailed errors. We also give parallel consistency and central limit results for correlated Gaussian noise as a benchmark. The asymptotic variance depends on the limiting aspect ratio, the true SNR, and a scalar noise-square fluctuation parameter. This explicit form yields feasible plug-in confidence intervals under independent noise in two cases where the fluctuation parameter can be estimated from response fourth moments: heterogeneous Gaussian noise and homogeneous non-Gaussian noise. Numerical simulations compare likelihood-based and method-of-moments confidence intervals under heterogeneous and non-Gaussian noise, and a real-data illustration demonstrates the resulting calibrations on high-dimensional text features.

math.ST↗

arXiv2025

Spatiotemporal spectral transfers in fluid dynamics

Motivated by previous work on kinetic energy cascades in the ocean and atmosphere, we develop a spatio-temporal spectral transfer tool that can be used to study scales of variability in generalized dynamical systems. In particular, we use generalized time-frequency methods from signal analysis to broaden the applicability of frequency transfers from theoretical to practical applications such as the study of ocean or atmosphere data or simulation output. We also show that triad interactions in wavenumber used to study kinetic energy and enstrophy cascades can be generalized to study triad interactions in frequency or wavenumber-frequency. We study the effects of sweeping on the locality of frequency transfers and frequency triad interactions to better understand the locality of spatio-temporal frequency transfers. As an illustrative example, we use the spatio-temporal spectral transfer tool to study the results of a simulation of two-dimensional homogeneous isotropic turbulence. This simulated fluid is forced at a well-defined wavenumber and frequency with dissipation occurring at both large and small scales, making this one of the first studies of "modulated turbulence" in two dimensions. Our results show that the spatio-temporal transfers we develop in this paper are robust to potential practical problems such as low sampling rates or nonstationarity in time series of interest. We anticipate that this method will be a useful tool in studying scales of spatio-temporal variability in a wide range of fluids applications as higher resolution data and simulations become more widely available.

physics.flu-dyn↗

arXiv2025

Efficient patient-centric EMR sharing block tree

Flexible sharing of electronic medical records (EMRs) is an urgent need in healthcare, as fragmented storage creates EMR management complexity for both practitioners and patients. Blockchain has emerged as a promising solution to address the limitations of centralized EMR systems regarding interoperability, data ownership, and trust concerns. Whilst its healthcare implementation continues to face scalability challenges, particularly in uploading lag time as EMR volumes increase. In this paper, we describe the design of a novel blockchain-based data structure, MedBlockTree, which aims to solve the scalability issue in blockchain-based EMR systems, particularly low block throughput and patient awareness. MedBlockTree leverages a chameleon hash function to generate collision blocks for existing patients and expand a single chain into a growing block tree with $n$ branches that are capable of processing $n$ new blocks in a single consensus round. We also introduce the EnhancedPro consensus algorithm to manage multiple branches and maintain network consistency. Our comprehensive simulation evaluates performance across four dimensions: branch number, worker number, collision rate, and network latency. Comparative analysis against a traditional blockchain-based EMR system demonstrates outstanding throughput improvements across all dimensions, achieving processing speeds $ν\cdot n$ times faster than conventional approaches.

cs.DC↗

arXiv2023

Iteratively Refined Behavior Regularization for Offline Reinforcement Learning

One of the fundamental challenges for offline reinforcement learning (RL) is ensuring robustness to data distribution. Whether the data originates from a near-optimal policy or not, we anticipate that an algorithm should demonstrate its ability to learn an effective control policy that seamlessly aligns with the inherent distribution of offline data. Unfortunately, behavior regularization, a simple yet effective offline RL algorithm, tends to struggle in this regard. In this paper, we propose a new algorithm that substantially enhances behavior-regularization based on conservative policy iteration. Our key observation is that by iteratively refining the reference policy used for behavior regularization, conservative policy update guarantees gradually improvement, while also implicitly avoiding querying out-of-sample actions to prevent catastrophic learning failures. We prove that in the tabular setting this algorithm is capable of learning the optimal policy covered by the offline dataset, commonly referred to as the in-sample optimal policy. We then explore several implementation details of the algorithm when function approximations are applied. The resulting algorithm is easy to implement, requiring only a few lines of code modification to existing methods. Experimental results on the D4RL benchmark indicate that our method outperforms previous state-of-the-art baselines in most tasks, clearly demonstrate its superiority over behavior regularization.

cs.LG↗

arXiv2022

Wall-modeled large-eddy simulation of three-dimensional turbulent boundary layer in a bent square duct

We conduct wall-modeled LES (WMLES) of a pressure-driven three-dimensional turbulent boundary layer (3DTBL) developing on the floor of a bent square duct to investigate the predictive capability of three widely used wall models, namely, a simple equilibrium stress model, an integral nonequilibrium model, and a PDE nonequilibrium model. The numerical results are compared with the experiment of Schwarz and Bradshaw (J. Fluid Mech. (1994), vol. 272, pp. 183-210). While the wall-stress magnitudes predicted by the three wall models are comparable, the PDE nonequilibrium wall model produces a substantially more accurate prediction of the wall-stress direction, followed by the integral nonequilibrium wall model. The wall-stress direction from the wall models is shown to have separable contributions from the equilibrium stress part and the integrated nonequilibrium effects, where how the latter is modeled differs among the wall models. The triangular plot of the wall-model solution reveals different capabilities of the wall models in representing variation of flow direction along the wall-normal direction. On the contrary, the outer LES solution is unaffected by the type of wall model used, resulting in nearly identical predictions of the mean and turbulent statistics in the outer region for all the wall models. This is explained by the vorticity dynamics and the inviscid skewing mechanism of generating the mean three-dimensionality. Finally, the LES solution in the outer layer is used to study the anisotropy of turbulence. In contrast to the canonical 2D wall turbulence, the Reynolds stress anisotropy exhibit strong non-monotonic behavior with increasing wall distance.

physics.flu-dyn↗