Bridging the Gap in ECG-Based Emotion Recognition: A Unified Evaluation of Deep Learning Models
该研究通过引入两个开源框架解决ECG情感识别中模型泛化能力评估的问题，使用深度学习模型进行跨数据集验证和比较分析。
0 citationsRead paper
Selected work
Representative Papers
Simultaneous inference of environmental and interaction forces in collective dynamics
研究通过非参数方法同时推断集体动态中的环境力和交互作用力，以提高模型准确性和预测能力。
0 citationsRead paper
Laziness of Quantum Walks on Graphs
本文研究了图上量子游走的惰性问题，通过平均混合矩阵的迹来衡量，并使用拉普拉斯量子游走工具确定了几类最惰性的连通图和树。
0 citationsRead paper
Generation of Synthetic Fingerphotos with GANs
This study addresses the scarcity of real-world data and privacy concerns in contactless fingerprint recognition by proposing a synthetic fingerprint generation framework based on StyleGAN2-ADA/3. Integrating biometric statistics with matching score analysis, this work presents the first quantitative assessment of the fidelity, privacy preservation, and diversity of synthetic samples. The research validates the efficacy of synthetic data for system development and establishes a standardized benchmark for quantitative evaluation. By releasing open-source code, this project provides a reliable solution that balances privacy protection with data augmentation for contactless biometrics, alongside a robust evaluation paradigm for future research in synthetic biometric data generation.
0 citationsRead paper
Dynamic E-unification
This work addresses the E-unification problem for non-ground terms in the presence of dynamically evolving sets of ground equations and inequations. It proposes a dynamic E-unification method that employs superposition calculus to saturate non-ground equations modulo a ground equational theory and introduces instantiation rules to match non-ground terms with ground terms, thereby enabling dynamic updates to the ground theory. To support this framework, the authors design a term ordering modulo theory that satisfies weak monotonicity and the subterm property, and they provide orientable inference rules for finite data structures—such as lists equipped with length and concatenation operations. The method is proven complete and capable of modeling evolving systems within quantified SMT problems.
0 citationsRead paper
Recent publications
Latest Papers
Bridging the Gap in ECG-Based Emotion Recognition: A Unified Evaluation of Deep Learning Models
该研究通过引入两个开源框架解决ECG情感识别中模型泛化能力评估的问题，使用深度学习模型进行跨数据集验证和比较分析。
0 citationsRead paper
Simultaneous inference of environmental and interaction forces in collective dynamics
研究通过非参数方法同时推断集体动态中的环境力和交互作用力，以提高模型准确性和预测能力。
0 citationsRead paper
Laziness of Quantum Walks on Graphs
本文研究了图上量子游走的惰性问题，通过平均混合矩阵的迹来衡量，并使用拉普拉斯量子游走工具确定了几类最惰性的连通图和树。
0 citationsRead paper
Generation of Synthetic Fingerphotos with GANs
This study addresses the scarcity of real-world data and privacy concerns in contactless fingerprint recognition by proposing a synthetic fingerprint generation framework based on StyleGAN2-ADA/3. Integrating biometric statistics with matching score analysis, this work presents the first quantitative assessment of the fidelity, privacy preservation, and diversity of synthetic samples. The research validates the efficacy of synthetic data for system development and establishes a standardized benchmark for quantitative evaluation. By releasing open-source code, this project provides a reliable solution that balances privacy protection with data augmentation for contactless biometrics, alongside a robust evaluation paradigm for future research in synthetic biometric data generation.
0 citationsRead paper
Dynamic E-unification
This work addresses the E-unification problem for non-ground terms in the presence of dynamically evolving sets of ground equations and inequations. It proposes a dynamic E-unification method that employs superposition calculus to saturate non-ground equations modulo a ground equational theory and introduces instantiation rules to match non-ground terms with ground terms, thereby enabling dynamic updates to the ground theory. To support this framework, the authors design a term ordering modulo theory that satisfies weak monotonicity and the subterm property, and they provide orientable inference rules for finite data structures—such as lists equipped with length and concatenation operations. The method is proven complete and capable of modeling evolving systems within quantified SMT problems.
0 citationsRead paper