Institution profile

Liaoning University

Academic institutionasia · cn
Official website
Research library6linked papers
Opportunities0open roles
Selected work

Representative Papers

SenCos-GEM: SENet-Calibrated and Law-of-Cosines-Constrained Geometry-Enhanced Molecular Representation for Property Prediction

Jul 14, 2026

This work addresses the limitations of existing 3D molecular representation methods, which lack explicit physical constraints, are susceptible to geometric noise, and neglect task-adaptive dynamic feature modulation—leading to negative transfer and catastrophic forgetting. To overcome these issues, we propose SenCos-GEM, a novel framework that introduces the law of cosines as a geometric consistency loss for the first time in molecular representation learning, thereby establishing a high-fidelity 3D spatial prior. Furthermore, SenCos-GEM explicitly decouples physical constraints from dynamic feature recalibration through a lightweight squeeze-and-excitation (SE) adapter integrated with a FiLM mechanism. Our method achieves new state-of-the-art results across multiple conformation-sensitive tasks on MoleculeNet, yielding relative error reductions of 12.9% RMSE on FreeSolv, 5.3% RMSE on Lipophilicity, and 8.2% MAE on QM9, while significantly enhancing discrimination of stereoisomers and robustness to conformational perturbations.

0 citationsRead paper

Graph Neural Networks for Generalized Mundlak Estimator under Network Confounding

May 27, 2026

This study addresses the limitations of traditional fixed-effects models, which rely on strong assumptions of linear additivity and independence, thereby struggling to accommodate group heterogeneity and within-group dependence and leading to biased cross-group comparisons. To overcome these issues, the authors propose a Graph Neural Network–based Generalized Mundlak Estimator (GME-GNN) that dispenses with conventional intercept terms and instead incorporates group-level balancing statistics to control for between-group confounding. By leveraging the message-passing mechanism of graph neural networks, the method adaptively learns nonlinear representations to flexibly capture intra-group interaction structures. The estimator is theoretically shown to possess double robustness and asymptotic normality. Both simulation experiments and empirical analyses demonstrate its superior performance over existing approaches in bias reduction and cross-group inference.

0 citationsRead paper

Uniform Agent-interpolation of Distributed Knowledge

Mar 01, 2026

This study addresses the uniform interpolation problem for propositional variables and agent symbols in epistemic logics with distributed knowledge—specifically, systems K, KD, and KT. Moving beyond traditional approaches that consider only propositional variables, this work is the first to incorporate agent symbols into the definition of uniform interpolation. Building on the sequent calculus framework of Murai and Sano (2020) and integrating Bílková’s (2007) interpolation technique, the authors devise a purely syntactic, algorithmic procedure for constructing interpolants. The paper establishes that these epistemic systems retain the uniform interpolation property even when extended with distributed knowledge, and it provides an effective, computable method for generating interpolating formulas, thereby significantly broadening the scope of interpolation theory in epistemic logic.

0 citationsRead paper
Recent publications

Latest Papers

SenCos-GEM: SENet-Calibrated and Law-of-Cosines-Constrained Geometry-Enhanced Molecular Representation for Property Prediction

Jul 14, 2026

This work addresses the limitations of existing 3D molecular representation methods, which lack explicit physical constraints, are susceptible to geometric noise, and neglect task-adaptive dynamic feature modulation—leading to negative transfer and catastrophic forgetting. To overcome these issues, we propose SenCos-GEM, a novel framework that introduces the law of cosines as a geometric consistency loss for the first time in molecular representation learning, thereby establishing a high-fidelity 3D spatial prior. Furthermore, SenCos-GEM explicitly decouples physical constraints from dynamic feature recalibration through a lightweight squeeze-and-excitation (SE) adapter integrated with a FiLM mechanism. Our method achieves new state-of-the-art results across multiple conformation-sensitive tasks on MoleculeNet, yielding relative error reductions of 12.9% RMSE on FreeSolv, 5.3% RMSE on Lipophilicity, and 8.2% MAE on QM9, while significantly enhancing discrimination of stereoisomers and robustness to conformational perturbations.

0 citationsRead paper

Graph Neural Networks for Generalized Mundlak Estimator under Network Confounding

May 27, 2026

This study addresses the limitations of traditional fixed-effects models, which rely on strong assumptions of linear additivity and independence, thereby struggling to accommodate group heterogeneity and within-group dependence and leading to biased cross-group comparisons. To overcome these issues, the authors propose a Graph Neural Network–based Generalized Mundlak Estimator (GME-GNN) that dispenses with conventional intercept terms and instead incorporates group-level balancing statistics to control for between-group confounding. By leveraging the message-passing mechanism of graph neural networks, the method adaptively learns nonlinear representations to flexibly capture intra-group interaction structures. The estimator is theoretically shown to possess double robustness and asymptotic normality. Both simulation experiments and empirical analyses demonstrate its superior performance over existing approaches in bias reduction and cross-group inference.

0 citationsRead paper

Uniform Agent-interpolation of Distributed Knowledge

Mar 01, 2026

This study addresses the uniform interpolation problem for propositional variables and agent symbols in epistemic logics with distributed knowledge—specifically, systems K, KD, and KT. Moving beyond traditional approaches that consider only propositional variables, this work is the first to incorporate agent symbols into the definition of uniform interpolation. Building on the sequent calculus framework of Murai and Sano (2020) and integrating Bílková’s (2007) interpolation technique, the authors devise a purely syntactic, algorithmic procedure for constructing interpolants. The paper establishes that these epistemic systems retain the uniform interpolation property even when extended with distributed knowledge, and it provides an effective, computable method for generating interpolating formulas, thereby significantly broadening the scope of interpolation theory in epistemic logic.

0 citationsRead paper