Using Genetic Algorithm to Optimize Information Dissemination

We formulate an optimal control problem to maximize the spread of a message at least campaigning cost. Information dissemination is captured using susceptible-infected (SI) epidemic process. The SI process is modeled using a system of ordinary differential equations. The standard model is modified to include effects of a control function. Then we formulate a cost function to account for both the cost of applying controls and reward due to spread of message. We show the existence of a solution for the formulated optimal control problem. Following this, the numerical solution is computed using the genetic algorithm. We show that genetic algorithm can be effectively used to solve the large scale optimal control problem formulated in this paper. The use of genetic algorithm technique is further justified by the fact that optimal control problems often have local minima. Genetic algorithm based techniques are more suited to handle such situations compared to the standard gradient descent methods which are likely to converge to one of the local minima.