15.08.2015 Views

Introduction to the Modeling and Analysis of Complex Systems

introduction-to-the-modeling-and-analysis-of-complex-systems-sayama-pdf

introduction-to-the-modeling-and-analysis-of-complex-systems-sayama-pdf

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

408CHAPTER 18. DYNAMICAL NETWORKS III: ANALYSIS OF NETWORK DYNAMICS<strong>the</strong> first two properties discussed above, <strong>the</strong> coefficient matrix <strong>of</strong> <strong>the</strong> equation has atleast one eigenvalue 0, <strong>and</strong> all <strong>the</strong> eigenvalues are 0 or negative. This means that <strong>the</strong>zero eigenvalue is actually <strong>the</strong> dominant one in this case, <strong>and</strong> its corresponding dominanteigenvec<strong>to</strong>r (h, or eigenspace, if <strong>the</strong> network is not connected) will tell us <strong>the</strong> asymp<strong>to</strong>ticstate <strong>of</strong> <strong>the</strong> network.Let’s review <strong>the</strong> o<strong>the</strong>r properties. The third property arises intuitively because, if<strong>the</strong> network is made <strong>of</strong> multiple connected components, each <strong>of</strong> those components behavesas a separate network <strong>and</strong> shows diffusion independently, converging <strong>to</strong> a differentasymp<strong>to</strong>tic value, <strong>and</strong> <strong>the</strong>refore, <strong>the</strong> asymp<strong>to</strong>tic state <strong>of</strong> <strong>the</strong> whole network should have asmany degrees <strong>of</strong> freedom as <strong>the</strong> connected components. This requires that <strong>the</strong> dominanteigenspace have as many dimensions, which is why <strong>the</strong>re should be as many degeneratedominant eigenvalue 0’s as <strong>the</strong> connected components in <strong>the</strong> network. The fourthproperty can be derived by combining this with <strong>the</strong> argument on property 1 above.And finally, <strong>the</strong> spectral gap. It is so called because <strong>the</strong> list <strong>of</strong> eigenvalues <strong>of</strong> a matrixis called a matrix spectrum in ma<strong>the</strong>matics. The spectral gap is <strong>the</strong> smallest non-zeroeigenvalue <strong>of</strong> L, which corresponds <strong>to</strong> <strong>the</strong> largest non-zero eigenvalue <strong>of</strong> −αL <strong>and</strong> thus<strong>to</strong> <strong>the</strong> mode <strong>of</strong> <strong>the</strong> network state that shows <strong>the</strong> slowest exponential decay over time. If <strong>the</strong>spectral gap is close <strong>to</strong> zero, this decay takes a very long time, resulting in slow diffusion.Or if <strong>the</strong> spectral gap is far above zero, <strong>the</strong> decay occurs quickly, <strong>and</strong> so does <strong>the</strong> diffusion.In this sense, <strong>the</strong> spectral gap <strong>of</strong> <strong>the</strong> Laplacian matrix captures some <strong>to</strong>pological aspects<strong>of</strong> <strong>the</strong> network, i.e., how well <strong>the</strong> nodes are connected <strong>to</strong> each o<strong>the</strong>r from a dynamicalviewpoint. The spectral gap <strong>of</strong> a connected graph (or, <strong>the</strong> second smallest eigenvalue <strong>of</strong>a Laplacian matrix in general) is called <strong>the</strong> algebraic connectivity <strong>of</strong> a network.Here is how <strong>to</strong> obtain a Laplacian matrix <strong>and</strong> a spectral gap in NetworkX:Code 18.1:>>> import networkx as nx>>> g = nx.karate_club_graph()>>> nx.laplacian_matrix(g)>>> nx.laplacian_spectrum(g)array([ 2.84494649e-15, 4.68525227e-01, 9.09247664e-01,1.12501072e+00, 1.25940411e+00, 1.59928308e+00,1.76189862e+00, 1.82605521e+00, 1.95505045e+00,2.00000000e+00, 2.00000000e+00, 2.00000000e+00,2.00000000e+00, 2.00000000e+00, 2.48709173e+00,

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!