Spectral approach to localize information spread in a network using Rao’s metaheuristic variants

Understanding different aspects of information spread have been a crucial direction of research in information and social sciences. Recent studies are found to be focused on accelerating and controlling the spread of information. One such study is to control the spread of information by manipulating the Principal Eigen Vector (PEV) of the network adjacency matrix. In this paper, Inverse Participation Ratio (IPR) is used to measure the localization of PEV. The studies say that maximum IPR leads to a high localization state of the network. In literature, IPR maximization is shown to be a hard problem. Therefore, we use approximation approach of solution to this problem by employing metaheuristic algorithms. In this direction, we propose graph versions of three Rao’s metaheuristic variants: Rao’s Algorithm, Jaya Algorithm, and Teaching–Learning-Based Optimization (TLBO) Algorithm for IPR maximization. Four evaluation parameters: Average Optimal IPR, Factor Improvement (IPR), Average Modification %, and Average Execution Time are used for comparison among the algorithms. The results show that TLBO outperforms the other two variants. In TLBO, the factor of improvement over the initial IPR value is found to be 1.5 in KC dataset with an average modification of 25%, 5.9 in GD 99 dataset with an average modification of 32%, 12.1 in GD 01 dataset with an average modification of 19%, 2.1 in USAL dataset with an average modification of 17%, and 15.1 in CL dataset with an average modification of 18%. Also, TLBO reports a better Average Optimal IPR value than the existing Random Perturbation (RP) approach.