Smallest eigenvalue of Laplacian of signed graphs

80 Views Asked by At

Given a weighted undirected graph $G=(V,E)$, which adjacency matrix is $A$. The weight of edges can be negative.

The graph laplacian is defined as

$l_{ij}=-a_{ij},\ \text{if } i\neq j$

$l_{ij}=\sum_{j=1,j\neq i}^na_{ij},\ \text{if } i=j$

Note that since this matrix may not enjoy the nice properties like positive semi-definiteness, and so on, the graph laplacian is usually modified as follows,

$l_{ij}=-a_{ij},\ \text{if } i\neq j$

$l_{ij}=\sum_{j=1,j\neq i}^n|a_{ij}|,\ \text{if } i=j$

see Spectral Analysis of Signed Graphs for Clustering, Prediction and Visualization

However, I am interested in the first definition of laplacian matrix. I am wondering is there any approach or work studying its spectrum, especially the smallest eigenvalue?

Thanks a lot for suggestions!