Extremum Inference Algorithm: A Clever Optimization Algorithm

Most intelligent algorithms can’t predict the existence and possible position of new better solution. This deficiency not only leads to the waste of computing resources, it also affects the efficiency to find the optimal solution. To fix this problem, this paper proposes a new algorithm named extremum inference algorithm. During the iterations, this algorithm uses high efficient algorithms to simultaneously searches for both local maxima and local minima, and then uses a simple principle to infer the existences and possible positions of unknown extremums. By doing this, the extremum inference algorithm can decide if there possibly exists new extremums and search for new extremums in the most possible regions. A strong version of extremum inference algorithm can even provide an approach to estimate searching accuracy, which is a measurement that can help to determine if the global optimal has been found. Our examples demonstrate that the extremum inference algorithm can efficiently solve multimodal problems which are harder to solve than other problems.