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

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

420CHAPTER 18. DYNAMICAL NETWORKS III: ANALYSIS OF NETWORK DYNAMICS18.6 Mean-Field Approximation on Scale-Free NetworksWhat if <strong>the</strong> network <strong>to</strong>pology is highly heterogeneous, like in scale-free networks, sothat <strong>the</strong> r<strong>and</strong>om network assumption is no longer applicable? A natural way <strong>to</strong> reconcilesuch heterogeneous <strong>to</strong>pology <strong>and</strong> mean-field approximation is <strong>to</strong> adopt a specific degreedistribution P (k). It is still a non-spatial summary <strong>of</strong> connectivities within <strong>the</strong> network, butyou can capture some heterogeneous aspects <strong>of</strong> <strong>the</strong> <strong>to</strong>pology in P (k).One additional complication <strong>the</strong> degree distribution brings in is that, because <strong>the</strong>nodes are now different from each o<strong>the</strong>r regarding <strong>the</strong>ir degrees, <strong>the</strong>y can also be differentfrom each o<strong>the</strong>r regarding <strong>the</strong>ir state distributions <strong>to</strong>o. In o<strong>the</strong>r words, it is nolonger reasonable <strong>to</strong> assume that we can represent <strong>the</strong> global state <strong>of</strong> <strong>the</strong> network by asingle “mean field” q. Instead, we will need <strong>to</strong> represent q as a function <strong>of</strong> degree k (i.e.,a bunch <strong>of</strong> mean fields, each for a specific k), because heavily connected nodes maybecome infected more <strong>of</strong>ten than poorly connected nodes do. So, here is <strong>the</strong> summary <strong>of</strong><strong>the</strong> new quantities that we need <strong>to</strong> include in <strong>the</strong> approximation:• P (k): Probability <strong>of</strong> nodes with degree k• q(k): Probability for a node with degree k <strong>to</strong> be infectedLet’s consider how <strong>to</strong> revise Table 18.1 using P (k) <strong>and</strong> q(k). It is obvious that all <strong>the</strong> q’sshould be replaced by q(k). It is also apparent that <strong>the</strong> third <strong>and</strong> fourth row (probabilities<strong>of</strong> transitions for currently infected states) won’t change, because <strong>the</strong>y are simply basedon <strong>the</strong> recovery probability p r . And <strong>the</strong> second row can be easily obtained once <strong>the</strong> firstrow is obtained. So, we can just focus on calculating <strong>the</strong> probability in <strong>the</strong> first row: Whatis <strong>the</strong> probability for a susceptible node with degree k <strong>to</strong> remain susceptible in <strong>the</strong> nexttime step?We can use <strong>the</strong> same strategy as what we did for <strong>the</strong> r<strong>and</strong>om network case. Thatis, we calculate <strong>the</strong> probability for <strong>the</strong> node <strong>to</strong> get infected from ano<strong>the</strong>r node, <strong>and</strong> <strong>the</strong>ncalculate one minus that probability raised <strong>to</strong> <strong>the</strong> power <strong>of</strong> <strong>the</strong> number <strong>of</strong> neighbors, <strong>to</strong>obtain <strong>the</strong> probability for <strong>the</strong> node <strong>to</strong> avoid any infection. For r<strong>and</strong>om networks, all o<strong>the</strong>rnodes were potential neighbors, so we had <strong>to</strong> raise (1 − p e qp i ) <strong>to</strong> <strong>the</strong> power <strong>of</strong> n − 1. Butthis is no longer necessary, because we are now calculating <strong>the</strong> probability for a nodewith <strong>the</strong> specific degree k. So, <strong>the</strong> probability that goes <strong>to</strong> <strong>the</strong> first row <strong>of</strong> <strong>the</strong> table shouldlook like:(1 − q(k)) (1 − something) k (18.29)Here, “something” is <strong>the</strong> joint probability <strong>of</strong> two events, that <strong>the</strong> neighbor node is infectedby <strong>the</strong> disease <strong>and</strong> that <strong>the</strong> disease is actually transmitted <strong>to</strong> <strong>the</strong> node in question. The

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

Saved successfully!

Ooh no, something went wrong!