The Power of Surrogate-Assisted Evolutionary Computing in Searching Vaccination Strategy

摘要

We propose to use genetic algorithms to search for the best vaccination strategy for a given scenario using the output of the simulation program as fitness score. The efficacy of vaccine varies significantly. Therefore, the real challenge is to find a good strategy without a priori knowledge of the efficacy of the vaccine. We use surrogate function instead of real simulation to achieve 1000 times speedup. The average of the absolute value of errors is less than 0.5% and the rank correlation coefficient is greater than 0.93 for almost all the scenarios. The optimal solution with surrogate has fitness value very close to one using simulation. The difference is generally less than one percent. Our search results confirm the convention wisdom to vaccinate school children first. It also reveals that there is appropriate strategy which works for most scenarios. It would be interesting to build autonomous software searches through the scenario space and adaptively revise the surrogate to produce better search results.