Notebooks
http://bactra.org/notebooks
Cosma's NotebooksenGraph Spectra
http://bactra.org/notebooks/2019/08/02#graph-spectra
<P>Yet Another Inadequate Placeholder: notes and references on the
eigenvalue and eigenvector spectra of graphs, or matrices derived
from graphs, and their applications, with links to
<a href="community-discovery.html">community discovery</a> being of particular
interest. (Since this is a new notebook, lots of the relevant references are
still scattered elsewhere.) A stub of a stab for explaining what it's about
follows.
<P>As you know, Babette, the spectrum of a matrix is the (multi-)set of its
eigenvalues; the spectral properties of a matrix refer to these eigenvalues and
associated eigenvectors. There are several matrices which can be defined for a
graph or network, and so several objects which could be considered the graph's
spectrum. The field of spectral graph theory as a whole, is mostly concerned
with the spectrum of the Laplacian matrix, which requires a little elucidation.
<P>In ordinary Euclidean space, when we say that some field "diffuses" over the
space, we mean that the total quantity is conserved, and the rate of change of
its density at any one point is proportional to the sum of the density's second
derivatives over space:
\[
\frac{\partial\rho}{\partial t}(x,y,z,t) = - \kappa \left( \frac{\partial^2 \rho}{\partial x^2}(x,y,z,t) + \frac{\partial^2 \rho}{\partial y^2}(x,y,z,t) + \frac{\partial^2 \rho}{\partial z^2}(x,y,z,t) \right)
\]
As Haldane said once, this is the mathematical way of saying "it oozes". The
sum of the squared second partial derivatives on the right hand side is the
"Laplacian" operator, written $\nabla^2$. This notion generalizes to other
manifolds, as the Laplace-Beltrami operator, and the corresponding equation
would express a conservative oozing along the manifold. It doesn't make sense
to take derivatives when space consists of nodes in a graph, but one can take
differences between neighboring points, and even second differences (like a
second derivative), and the matrix which does is the graph's Laplacian. In
symbols, if we write \( A \) for the adjacency matrix and \( D \) for the
diagonal matrix giving each node's degree, then the Laplacian of an
(undirected) graph is
\[
L = D^{-1}(D-A)
\]
and this is the <a href="markov.html">generator</a> of a continuous-time random
walk on the graph. Alternately, one considers the continuous-time diffusion
equation,
\[
\frac{d\rho}{dt} = -\kappa L \rho
\]
or its discrete-time analog,
\[
\rho_{t+h} = (I-h\kappa L)\rho_t
\]
(Physically, the factor $\kappa$ expresses how fast the diffusion happens in
real, clock time. By re-scaling time, however, one can always set $\kappa$ or
$h\kappa$ to 1, so I'll do so.)
<P>The conservation of the field $\rho$ is expressed by the fact that a
constant vector, say the all-ones vector, is always an eigenvector of $L$, with
eigenvalue $0$. (Equivalently, $I-L$'s leading eigenvalue is always $1$, and
it, too, has the all-ones vector as an eigenvector.) The number of zero
eigenvalues counts the number of connected components of the graph. (The
eigenvectors are piece-wise constant on each component.) To keep things
simple, I'll assume that there is only one connected component.
<P>All of the eigenvalues of $L$ are non-negative, so we can number them
$0=\lambda_0 < \lambda_1 \leq \lambda_2 \ldots \leq \lambda_n$. What do we
learn from the rest of the eigen-spectrum? The eignevector with the first
positive eigenvalue has both positive and negative entries (because it's
orthogonal to the all-ones vector), and in fact divides the graph into two
parts. More specifically, it divides the graph into two parts across which
diffusion is very slow. To see this, consider an eigenvector expansion of some arbitrary initial state:
\[
\rho_0 = \sum_{i=0}^{n-1}{a_i v_i}
\]
Now we evolve it forward in time:
\[
\rho_t = \sum_{i=0}^{n-1}{a_i (I-L)^t v_i}
\]
Notice that every eigenvector of $L$ is also an eigenvector of $I-L$, with a simple relation between the eigenvalues:
\[
(I-L) v_i = v_i - Lv_i = v_i - \lambda_i v_i = (1-\lambda_i) v_i
\]
So
\[
\rho_t = \sum_{i=0}^{n-1}{a_i (1-\lambda_i)^t v_i}
\]
As $t \rightarrow \infty$, diffusion washes away any inequalities in the field,
because the contribution of every eigenvector except $ v_0 $ , the all-ones
vector, gets exponentially suppressed $(1-\lambda_i)^t \rightarrow 0$ . But the
slowest decay mode is $ v_1 $ --- that's the pattern over the graph which is
least affected by diffusion, and survives the longest.
<P>See also:
<a href="network-data-analysis.html">Analysis of network data</a>;
<a href="complex-networks.html">Complex networks</a>;
<a href="epidemic-models.html">Epidemic models</a>;
<a href="graph-limits.html">Graph limits</a>;
<a href="graph-theory.html">Graph theory</a> (general);
<a href="math.html">Math I ought to learn</a>;
<a href="topology-and-synchronization.html">Topology and Synchronization</a>
<ul>Recommended, big picture:
<li>Fan R. K. Chung, <cite><a href="../weblog/algae-2014-12.html#chung">Spectral Graph Theory</a></cite>
</ul>
<ul>Recommended, close-ups:
<li>M. E. J. Newman, "Finding community structure in networks using the
eigenvectors of
matrices", <a href="http://dx.doi.org/10.1103/PhysRevE.74.036104"><cite>Physical
Review E</cite> <strong>74</strong> (2006):
036104</a>, <a href="http://arxiv.org/abs/physics/0605087">physics/0605087</a>
<li>Jesse Shore and Benjamin Lubin, "Spectral goodness of fit
for network
models", <a href="http://dx.doi.org/10.1016/j.socnet.2015.04.004"><cite>Social
Networks</cite> <strong>43</strong> (2015): 16–-27</a>, <a href="http://arxiv.org/abs/1407.7247">arxiv:1407.7247</a>
<li>Laura M. Smith, Kristina Lerman, Cristina Garcia-Cardona, Allon G. Percus, Rumi Ghosh, "Spectral Clustering with Epidemic Diffusion",
<a href="http://dx.doi.org/10.1103/PhysRevE.88.042813"><cite>Physical Review E</cite> <strong>88</strong>
</ul>
<ul>To read:
<li>Fan R. K. Chung, "From Quasirandom graphs to Graph Limits and Graphlets", <a href="http://arxiv.org/abs/1203.2269">arxiv:1203.2269</a> [Formerly titled
"Graphlets: a Spectral Perspective for Graph Limits"]
<li>M. A. M. de Aguiar and Y. Bar-Yam, "Spectral Analysis and the
Dynamic Response of Complex
Networks", <a href="http://arxiv.org/abs/nlin.AO/0306043">nlin.AO/0306043</a>
<li>Sanjeev Chauhan, Michelle Girvan and Edward Ott, "Spectral properties of networks with community structure",
<a href="http://dx.doi.org/10.1103/PhysRevE.80.056114"><cite>Physical Review E</cite> <strong>80</strong>
(2009): 056114</a>
<li>Iles Farkas, I. Derenyi, H. Jeong, Z. Neda, Z. N. Oltvai,
E. Ravasz, A. Schubert, Albert-Laszlo Barabasi and Tamas Vicsek, "Networks in
life: Scaling properties and eigenvalue spectra," <a
href="http://arxiv.org/abs/cond-mat/0303106">cond-mat/0303106</a>
<li>Kwang-Il Goh, B. Kahng, and D. Kim, "Spectra and eigenvectors of
scale-free networks,"
<a href="http://arxiv.org/abs/cond-mat/0103337">cond-mat/0103337</a>
<li>U Kang, Brendan Meeder, Evangelos E. Papalexakis, and Christos Faloutsos, "HEigen: Spectral Analysis for Billion-Scale Graphs" [<a href="http://www.cs.cmu.edu/~ukang/papers/HeigenTKDE2012.pdf">PDF preprint</a>]
<li>Can M. Le, Roman Vershynin, "Concentration and regularization of random graphs", <a href="http://arxiv.org/abs/1506.00669">arxiv:1506.00669</a>
<li>Raj Rao Nadakuditi and M. E. J. Newman, "Spectra of random graphs with arbitrary expected degrees", <a href="http://dx.doi.org/"><cite>Physical Review E</cite> <strong>87</strong> (2013): 012803</a>, <a href="http://arxiv.org/abs/1306.2507">arxiv:1306.2507</a>
<li>M. E. J. Newman, "Spectra of networks containing short loops",
<a href="http://dx.doi.org/10.1103/PhysRevE.100.012314"><cite>Physical Review E</cite> <strong>100</strong> (2019): 012314</a>
<li>Géza Ódor, "Spectral analysis and slow spreading dynamics on complex networks", <a href="http://dx.doi.org/10.1103/PhysRevE.88.032109"><cite>Physical Review E</cite> <strong>88</strong> (2013): 032109</a>,
<a href="http://arxiv.org/abs/1306.3401">arxiv:1306.3401</a>
<li>Ulrike von Luxburg, Mikhail Belkin, Olivier Bousquet, "Consistency of spectral clustering", <a href="http://arxiv.org/abs/0804.0678">arxiv:0804.0678</a>
</ul>