Abstract Network construction has been heavily exploited in multivariate data analysis. In many cases, connections between a large portion of variables are of minimal importance. As such, partial graphs have played an important role in network construction. Due to the existence of a multiplicative normalization constant, the existing construction approaches may bear high computational cost. To reduce the computational complexity, the conditional score matching for high-dimensional partial graphical models is proposed. This approach is uniquely advantageous by being not influenced by the multiplicative normalization constant. An effective computational algorithm is developed, and it is shown that the computational complexity of the proposed method is less than that of those in the literature. Statistical properties are established, and two extensions are explored to incorporate more information and accommodate more general distributions. A wide spectrum of simulations and the analysis of a breast cancer gene expression dataset demonstrate competitive performance of the proposed methods.

中文翻译：

---

高维部分图形模型的条件得分匹配

摘要 网络构建在多元数据分析中得到了大量利用。在许多情况下，大部分变量之间的联系并不重要。因此，部分图在网络构建中发挥了重要作用。由于乘法归一化常数的存在，现有的构造方法可能会承担很高的计算成本。为了降低计算复杂度，提出了高维局部图模型的条件分数匹配。这种方法的独特优势在于不受乘法归一化常数的影响。开发了一种有效的计算算法，结果表明该方法的计算复杂度低于文献中的计算复杂度。统计属性成立，并且探索了两个扩展以包含更多信息并适应更一般的分布。广泛的模拟和乳腺癌基因表达数据集的分析证明了所提出方法的竞争性能。