Penalized likelihood estimation of probability density functions using compositional splines
本文提出了一种基于中心对数比变换和ZB样条基函数的惩罚最大似然框架，直接从原始观测数据估计概率密度函数，减少两阶段方法带来的偏差。
0 citationsRead paper
Selected work
Representative Papers
Compositional Periodic Spline Approximation for Circular Density Data in Bayes Spaces
This work addresses the challenge of modeling circular density data characterized by periodicity and relative structure by proposing a novel periodic spline approach within the Bayes space framework. By applying the centered log-ratio transformation, densities are mapped into an L² subspace subject to a zero-integral constraint, enabling the construction of spline bases that simultaneously respect periodicity and Hilbert space structure. The method unifies smoothing and penalized spline estimation in a matrix formulation for computational efficiency. It represents the first integration of periodic splines with Bayes space theory, preserving the relative nature and interpretability of densities while facilitating subsequent functional data analysis. Experiments on wind direction data demonstrate that the proposed approach yields smooth, plausible, and interpretable density estimates, offering a new paradigm for modeling complex circular density data.
0 citationsRead paper
Exploring Urban Land Use Patterns by Pattern Mining and Unsupervised Learning
本文通过模式挖掘和无监督学习方法，利用Urban Atlas 2018数据识别欧洲城市中重复出现的土地使用配置，采用Ward聚类法进行分析。
0 citationsRead paper
Completeness of Relational Algebra via Cylindric Algebra
This study investigates the completeness of relational algebra with respect to range-restricted first-order logic formulas. By embedding relational algebra within the framework of cylindric algebras, the work proposes a novel algebraic approach to proving completeness, circumventing the limitations inherent in traditional model-theoretic methods. Building on this theoretical foundation, the authors design and implement an effective algorithm capable of automatically translating any range-restricted first-order formula into an equivalent relational algebra expression. This contribution not only furnishes an alternative formal proof of the completeness of relational algebra but also establishes a generalizable algebraic basis for future extensions to relational models handling incomplete or uncertain information.
0 citationsRead paper
GreCon3: Mitigating High Resource Utilization of GreCon Algorithms for Boolean Matrix Factorization
This work proposes GreCon3, an improved algorithm for Boolean matrix factorization based on formal concept analysis, addressing the high memory consumption and low computational efficiency of GreCon and GreCon2 when applied to large-scale binary datasets. GreCon3 introduces a space-efficient data structure and an incremental initialization strategy to enhance the tracking of uncovered data entries, eliminates irrelevant terms, and refines the initial factor extraction process to reduce redundant computations. Experimental results demonstrate that GreCon3 substantially reduces memory usage and accelerates the decomposition process, enabling the successful handling of large-scale binary datasets previously intractable with earlier methods. This advancement significantly improves the scalability of formal concept analysis–based Boolean matrix factorization.
0 citationsRead paper
Recent publications
Latest Papers
Penalized likelihood estimation of probability density functions using compositional splines
本文提出了一种基于中心对数比变换和ZB样条基函数的惩罚最大似然框架，直接从原始观测数据估计概率密度函数，减少两阶段方法带来的偏差。
0 citationsRead paper
Compositional Periodic Spline Approximation for Circular Density Data in Bayes Spaces
This work addresses the challenge of modeling circular density data characterized by periodicity and relative structure by proposing a novel periodic spline approach within the Bayes space framework. By applying the centered log-ratio transformation, densities are mapped into an L² subspace subject to a zero-integral constraint, enabling the construction of spline bases that simultaneously respect periodicity and Hilbert space structure. The method unifies smoothing and penalized spline estimation in a matrix formulation for computational efficiency. It represents the first integration of periodic splines with Bayes space theory, preserving the relative nature and interpretability of densities while facilitating subsequent functional data analysis. Experiments on wind direction data demonstrate that the proposed approach yields smooth, plausible, and interpretable density estimates, offering a new paradigm for modeling complex circular density data.
0 citationsRead paper
Exploring Urban Land Use Patterns by Pattern Mining and Unsupervised Learning
本文通过模式挖掘和无监督学习方法，利用Urban Atlas 2018数据识别欧洲城市中重复出现的土地使用配置，采用Ward聚类法进行分析。
0 citationsRead paper
Completeness of Relational Algebra via Cylindric Algebra
This study investigates the completeness of relational algebra with respect to range-restricted first-order logic formulas. By embedding relational algebra within the framework of cylindric algebras, the work proposes a novel algebraic approach to proving completeness, circumventing the limitations inherent in traditional model-theoretic methods. Building on this theoretical foundation, the authors design and implement an effective algorithm capable of automatically translating any range-restricted first-order formula into an equivalent relational algebra expression. This contribution not only furnishes an alternative formal proof of the completeness of relational algebra but also establishes a generalizable algebraic basis for future extensions to relational models handling incomplete or uncertain information.
0 citationsRead paper
GreCon3: Mitigating High Resource Utilization of GreCon Algorithms for Boolean Matrix Factorization
This work proposes GreCon3, an improved algorithm for Boolean matrix factorization based on formal concept analysis, addressing the high memory consumption and low computational efficiency of GreCon and GreCon2 when applied to large-scale binary datasets. GreCon3 introduces a space-efficient data structure and an incremental initialization strategy to enhance the tracking of uncovered data entries, eliminates irrelevant terms, and refines the initial factor extraction process to reduce redundant computations. Experimental results demonstrate that GreCon3 substantially reduces memory usage and accelerates the decomposition process, enabling the successful handling of large-scale binary datasets previously intractable with earlier methods. This advancement significantly improves the scalability of formal concept analysis–based Boolean matrix factorization.
0 citationsRead paper