Robustness and sampling enhancement in random geometric network

Abstract

We investigate how much information about a large network can be retained in the sub-network samples. In particular, we subject a random geometric network, G(R, Ï), to varying sampling conditions such as type of sampling and fractional sample size, α. We test the true network and its samples to stochastic attack and measure how robustness is maintained in the samples. We correlate robustness to a cross-over point p_c that marks the onset of fragmentation in a fully connected network. We also identify the critical fractional sample size α_c as the point of optimal information retrieved about the robustness of the sub-network samples. The results show that by increasing the connection radius parameter R, robustness can be enhanced and critical fractional sample size can be reduced significantly. The reduction of the minimum fractional sample size needed for optimal information about the true network is an important finding of the study. This will allow us to reduce data needed in characterizing very large real world networks.