CEU Electronic Theses and Dissertations, 2016
| Author | Rózemberczki, Benedek András |
|---|---|
| Title | Homophily Rearrangement Algorithms and Similarity Based Diffusion on Networks |
| Summary | The thesis first proposes three novel algorithms that can rearrange generic vertex features of a network in a way that target levels of homophily or heterophily regarding the variables are achievable. The thesis generalizes these algorithms to multivariate networks where vertices have more than a single generic vertex feature. Simulation results demonstrate that the homophily rearrangement algorithms' expected convergence time to a target homophily level depends on the network size, the target homophily level and in multivariate cases on the correlation among the generic vertex features. In addition, I extend the susceptible-infect model in a way that the transmission probability between vertices depends on the dissimilarity of generic vertex features. With the ability to control the level of homophily on the network with homophily rearrangement networks, the properties of the proposed similarity based diffusion model can be investigated. Empirical results establish the phenomenon that homophily propagates similarity based diffusion. Moreover, I demonstrate that the opposite is true when agents discriminate based on breed and the initial seeding originates from the discriminating breed. |
| Supervisor | Mantegna, Rosario Nunzio |
| Department | Economics MA |
| Full text | https://www.etd.ceu.edu/2016/rozemberczki_benedek.pdf |
Visit the CEU Library.
© 2007-2021, Central European University