5
$\begingroup$

Need some help / ideas to proceed. Stuck for a while on this.

In the literature of epidemic theory, it is found that the epidemic threshold is $1/\lambda_{\max}(A)$ where $\lambda_{\max}(A)$ is the largest eigenvalue of the adjacency matrix $A$, thus implying that the outbreak is basically dependent on the spectral radius of the underlying contact graph.

Say, now, there's a new kind of virus that spreads by a different mechanism (i.e., no longer purely based on the existence of a link between two nodes). For instance, say each time it infects a specific type of people (nodes) (e.g., male/female, young/old, Asian/European/American/African etc.) and then it mutates and infect different kind of people (nodes) next time instance.

For this, I come up with a matrix, say $X$, that records the probability of the virus infecting each node (i.e., $x_{ij}$ = probability of node $j$ being infected if $i$ is the infectant). It is a non-symmetric square hollow matrix with positive values for all off-diagonal elements in the matrix.

With this infection probability matrix $X$ in mind, I'm struggling on how to derive the epidemic threshold. Is it still $1/\lambda_{\max}(X)$ by analogy of the conventional epidemic theory? Any idea, suggestions, hints or pointers will be appreciated.

Secondly, if I have two of such matrices $X_1$ and $X_2$ essentially describing the behaviour of two viruses, how do I compare them using their corresponding probability matrix, $X$?

Thank you in advance.

$\endgroup$
2
  • $\begingroup$ Interesting question, but can you clarify: by "hollow matrix", do you mean one whose diagonal entries are all zero? Wikipedia gives two different meanings for the term: en.wikipedia.org/wiki/Hollow_matrix $\endgroup$ Jan 16, 2014 at 13:11
  • $\begingroup$ Apology. By hollow matrix, I'm referring to the matrix which has zero for all it's diagonal elements ($tr(X)=0$). $\endgroup$
    – Val K
    Jan 16, 2014 at 13:29

1 Answer 1

2
$\begingroup$

The problem is addressed in

Peng, Chengbin, Xiaogang Jin, and Meixia Shi. "Epidemic threshold and immunization on generalized networks." Physica A: Statistical Mechanics and its Applications 389.3 (2010): 549-560. Link

An epidemic will become extinct if and only if the spectral radius of matrix $X$ is smaller than 1.

$\endgroup$
2
  • $\begingroup$ Thank you very much for this pointer which seems very relevant indeed. One question though: It seems that the system matrix $M$ mentioned in this paper seems to be symmetric but my probability matrix $X$ is not (in fact, it has complex eigenvalues as well). Does this invalidate the theorem derived in that paper? Thank you again. $\endgroup$
    – Val K
    Jan 16, 2014 at 15:15
  • $\begingroup$ @ValK The authors do not assume that the matrix is symmetric - see e.g. Section "Solution for (17) with nonsymmetric PAM" $\endgroup$
    – Waldemar
    Jan 16, 2014 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.