A matrix approach for finding extrema: problems with modularity, hierarchy, and overlap
A matrix approach for finding extrema: problems with modularity, hierarchy, and overlap
| Hi-index | 0.00 |
We propose an optimization algorithm in which a factor graph is used to encode the underlying distribution of the problem. The factor graph is learned using symmetric non-negative matrix factorization (SNMF) approach and bivariate statistics. Based on experimental and theoretical discussions, the proposed approach is capable of solving the optimization problems in polynomial time with polynomial number of evaluations.