Chung Fan, Lu Linyuan, Vu Van
Department of Mathematics, University of California at San Diego, La Jolla 92093-0112, USA.
Proc Natl Acad Sci U S A. 2003 May 27;100(11):6313-8. doi: 10.1073/pnas.0937490100. Epub 2003 May 12.
In the study of the spectra of power-law graphs, there are basically two competing approaches. One is to prove analogues of Wigner's semicircle law, whereas the other predicts that the eigenvalues follow a power-law distribution. Although the semicircle law and the power law have nothing in common, we will show that both approaches are essentially correct if one considers the appropriate matrices. We will prove that (under certain mild conditions) the eigenvalues of the (normalized) Laplacian of a random power-law graph follow the semicircle law, whereas the spectrum of the adjacency matrix of a power-law graph obeys the power law. Our results are based on the analysis of random graphs with given expected degrees and their relations to several key invariants. Of interest are a number of (new) values for the exponent beta, where phase transitions for eigenvalue distributions occur. The spectrum distributions have direct implications to numerous graph algorithms such as, for example, randomized algorithms that involve rapidly mixing Markov chains.
在幂律图的光谱研究中,基本上有两种相互竞争的方法。一种是证明维格纳半圆定律的类似物,而另一种则预测特征值遵循幂律分布。尽管半圆定律和幂律毫无共同之处,但我们将表明,如果考虑适当的矩阵,这两种方法本质上都是正确的。我们将证明(在某些温和条件下)随机幂律图的(归一化)拉普拉斯算子的特征值遵循半圆定律,而幂律图的邻接矩阵的谱遵循幂律。我们的结果基于对具有给定期望度的随机图及其与几个关键不变量的关系的分析。有趣的是,对于指数β有一些(新的)值,在这些值处会发生特征值分布的相变。谱分布对许多图算法有直接影响,例如涉及快速混合马尔可夫链的随机算法。