arXiv subjects

Juan Shen

Publications and source records attributed to Juan Shen.

9 recordsLinked to original sources

arXiv2026

A Unified Adaptive Enrichment Design for Power Enhancement

Randomized controlled trials (RCTs) are the gold standard for evaluating treatment effects, but fixed eligibility criteria and enrollment decisions can be inefficient, especially when treatment effects vary across patient subpopulations. Adaptive enrichment trials update enrollment using interim data to improve efficiency. Enrichment methods are developed for two settings: prespecified subgroups, and continuous covariates where enrollment is guided by a learned cutoff. Many enrichment designs adopt discontinuous rules that favor one single subgroup, which may induce "winner's curse" bias if final estimation does not account for the data-dependent enrollment decision and require additional bias correction. We propose a unified framework that bridges these settings by formulating enrichment as a regularized optimization over the enrolled covariate distribution. In a two-stage design, Stage 2 selects an enrollment mixture by maximizing a power objective while penalizing deviation from a prespecified baseline target population through a Kullback-Leibler divergence term, providing a smooth alternative to pick-the-winner rules; the same formulation extends naturally to optimizing enrollment over continuous covariates. The resulting estimand is the average treatment effect in the trial population induced by the data-adaptive enrollment rule, so uncertainty quantification must account for randomness in learning the optimal enrollment rule, in addition to outcome estimation. We derive an influence-function representation for the estimated optimal enrollment rule and account for it in the final estimator, yielding an explicit asymptotic variance decomposition into decision uncertainty and outcome-estimation uncertainty. Simulations demonstrate improved power relative to conventional enrichment approaches while substantially reducing winner's curse bias in treatment effect estimation.

stat.ME↗

arXiv2026

Bayesian Environment Invariant Regression

The availability of data from multiple heterogeneous environments has motivated methods that remain reliable under distributional shifts. When the joint distribution of response and predictors varies across environments, the response may still depend on a subset of predictors through an invariant mechanism. Existing methods typically assess candidate invariant sets through pooled stability criteria, treating environmental variation as nuisance. In this paper, we propose a Bayesian framework that explicitly separates a shared response mechanism from environment-specific or response-dependent associations, exploiting heterogeneity as evidence for structure learning. A competitive spike-and-slab prior is designed to force each predictor to compete between invariant and non-invariant spurious effects. Under a tractable working model, we establish invariant model selection consistency and posterior contraction for invariant coefficients. We further study the presence of irrelevant predictors, characterize posterior concentration on an equivalent invariant class, and introduce a post-selection refinement that consistently recovers the minimal invariant model. Simulations and a real application illustrate the robustness and finite-sample efficiency of the proposed method.

stat.ME↗

arXiv2025

Syntax-Guided Diffusion Language Models with User-Integrated Personalization

Large language models have made revolutionary progress in generating human-like text, yet their outputs often tend to be generic, exhibiting insufficient structural diversity, which limits personalized expression. Recent advances in diffusion models have opened new opportunities for improving language generation beyond the limitations of autoregressive paradigms. In this work, we propose a syntax-guided diffusion language model that integrates structural supervision and personalized conditioning to enhance text quality, diversity, and controllability. We introduce a cascaded framework that generates syntactic guidance before conditional text generation, and further generalize it to a novel noncascaded architecture for better alignment between structure and content. By incorporating syntactic information in the generating process, the proposed model better captures the lexical and structural characteristics of stylistic sentence construction. To enable fine-grained personalization, we develop a shared representation mechanism that facilitates information integration across users, supporting both faithful stylistic generation and generalizable zero-shot inference. Extensive experiments on multiple tasks demonstrate the superiority of our approach in fluency, diversity, and stylistic fidelity. Further qualitative analyses highlight its interpretability and flexibility in learning personalized patterns.

cs.CL↗

arXiv2024

Covariate-Elaborated Robust Partial Information Transfer with Conditional Spike-and-Slab Prior

The popularity of transfer learning stems from the fact that it can borrow information from useful auxiliary datasets. Existing statistical transfer learning methods usually adopt a global similarity measure between the source data and the target data, which may lead to inefficiency when only partial information is shared. In this paper, we propose a novel Bayesian transfer learning method named ``CONCERT'' to allow robust partial information transfer for high-dimensional data analysis. A conditional spike-and-slab prior is introduced in the joint distribution of target and source parameters for information transfer. By incorporating covariate-specific priors, we can characterize partial similarities and integrate source information collaboratively to improve the performance on the target. In contrast to existing work, the CONCERT is a one-step procedure which achieves variable selection and information transfer simultaneously. We establish variable selection consistency, as well as estimation and prediction error bounds for CONCERT. Our theory demonstrates the covariate-specific benefit of transfer learning. To ensure the scalability of the algorithm, we adopt the variational Bayes framework to facilitate implementation. Extensive experiments and two real data applications showcase the validity and advantages of CONCERT over existing cutting-edge transfer learning methods.

cs.LG↗

arXiv2014

Localized versus itinerant states created by multiple oxygen vacancies in SrTiO3

Oxygen vacancies in strontium titanate surfaces (SrTiO3) have been linked to the presence of a two-dimensional electron gas with unique behavior. We perform a detailed density functional theory study of the lattice and electronic structure of SrTiO3 slabs with multiple oxygen vacancies, with a main focus on two vacancies near a titanium dioxide terminated SrTiO3 surface. We conclude based on total energies that the two vacancies preferably inhabit the first two layers, i.e. they cluster vertically, while in the direction parallel to the surface, the vacancies show a weak tendency towards equal spacing. Analysis of the nonmagnetic electronic structure indicates that oxygen defects in the surface TiO2 layer lead to population of Ti t2g states and thus itinerancy of the electrons donated by the oxygen vacancy. In contrast, electrons from subsurface oxygen vacancies populate Ti eg states and remain localized on the two Ti ions neighboring the vacancy. We find that both, the formation of a bound oxygen-vacancy state composed of hybridized Ti 3eg and 4p states neighboring the oxygen vacancy as well as the elastic deformation after extracting oxygen contribute to the stabilization of the in-gap state.

cond-mat.str-el↗

arXiv2013

A valence bond liquid on the honeycomb lattice

The honeycomb lattice material Li2RuO3 undergoes a dimerization of Ru4+ cations on cooling below 270C, where the magnetic susceptibility vanishes. We use density functional theory calculations to show that this reflects the formation of a 'valence bond crystal', with a strong bond disproportionation. On warming, x-ray diffraction shows that discrete three-fold symmetry is regained on average, and the dimerization apparently disappears. In contrast, local structural measurements using high-energy x-rays, show that disordered dimers survive at the nanoscale up to at least 650C. The high temperature phase of Li2RuO3 is thus an example of a valence bond liquid, where thermal fluctuations drive resonance between different dimer coverages, a classic analogue of the resonating valence bond state often discussed in connection with high T$_c$ cuprates.

cond-mat.str-el↗

arXiv2012

Ab initio study of the two-dimensional metallic state at the surface of SrTiO3: importance of oxygen vacancies

Motivated by recent angle-resolved photoemission spectroscopy (ARPES) observations of a highly metallic two-dimensional electron gas (2DEG) at the (001) vacuum-cleaved surface of SrTiO3 and the subsequent discussion on the possible role of oxygen vacancies for the appearance of such a state (Ref 1), we analyze by means of density functional theory (DFT) the electronic structure of various oxygen-deficient SrTiO3 surface slabs. We find a significant surface reconstruction after introducing oxygen vacancies and we show that the charges resulting from surface-localized oxygen vacancies --independently of the oxygen concentration-- redistribute in the surface region and deplete rapidly within a few layers from the surface suggesting the formation of a 2DEG. We discuss the underlying model emerging from such observations.

cond-mat.mtrl-sci↗

arXiv2012

Physisorption of an organometallic platinum complex on silica. An ab initio study

The interaction of trimethyl methylcyclopentadienyl platinum (MeCpPtMe3) with a fully hydroxylated SiO2 surface has been explored by means of ab initio calculations. A large slab model cut out from the hydroxylated beta-cristobalite SiO2 (111) surface was chosen to simulate a silica surface. Density functional theory calculations were performed to evaluate the energies of MeCpPtMe3 adsorption to the SiO2 surface. Our results show that the physisorption of the molecule is dependent on both (i) the orientation of the adsorbate and (ii) the adsorption site on the substrate. The most stable configuration was found with the MeCp and Me3 groups of the molecule oriented towards the surface. Finally, we observe that van-der-Waals corrections are crucial for the stabilization of the molecule on the surface. We discuss the relevance of our results for the growth of Pt-based nanostructured materials via deposition processes such as electron beam induced deposition.

cond-mat.mtrl-sci↗

arXiv2011

Interaction of W(CO)$_6$ with SiO$_2$ Surfaces -- A Density Functional Study

The interaction of tungsten hexacarbonyl W(CO)$_6$ precursor molecules with SiO$_2$ substrates is investigated by means of density functional theory calculations with and without inclusion of long range van der Waals interactions. We consider two different surface models, a fully hydroxylated and a partially hydroxylated SiO$_2$ surface, corresponding to substrates under different experimental conditions. For the fully hydroxylated surface we observe only a weak interaction between the precursor molecule and the substrate with physisorption of W(CO)$_6$. Inclusion of van der Waals corrections results in a stabilization of the molecules on this surface, but does not lead to significant changes in the chemical bonding. In contrast, we find a spontaneous dissociation of the precursor molecule on the partially hydroxylated SiO$_2$ surface where chemisorption of a W(CO)$_5$ fragment is observed upon removal of one of the CO ligands from the precursor molecule. Irrespective of the hydroxylation, the precursor molecule prefers binding of more than one of its CO ligands. In the light of these results, implications for the initial growth stage of tungsten nano-deposits on SiO$_2$ in an electron beam induced deposition process are discussed.

cond-mat.mtrl-sci↗