Brussels, Belgium Belgium
Dec. 10, 2012 to Dec. 10, 2012
Recent research has focused on sampling online social networks (OSNs) using traditional Markov Chain Monte Carlo (MCMC) techniques such as the Metropolis-Hastings algorithm (MH). While these methods have exhibited some success, the techniques suffer from slow mixing rates by themselves, and the resulting sample is usually approximate. An appealing solution is to apply the state of the art MCMC technique, Coupling From The Past (CFTP), for perfect sampling of OSNs. In this initial research, we explore theoretical and methodological issues such as customizing the update function and generating small sets of non-trivial states to adapt CFTP for sampling OSNs. Our research proposes the possibility of achieving perfect samples from large and complex OSNs using CFTP.
Couplings, Markov processes, Standards, Convergence, Facebook, Monte Carlo methods, Coupling From The Past, Sampling, Online Social Networks, Markov Chain Monte Carlo
Kenton White, Guichong Li, Nathalie Japkowicz, "Sampling Online Social Networks Using Coupling from the Past", ICDMW, 2012, 2013 IEEE 13th International Conference on Data Mining Workshops, 2013 IEEE 13th International Conference on Data Mining Workshops 2012, pp. 266-272, doi:10.1109/ICDMW.2012.126