An interactive evolutionary algorithm with adaptive evaluation mechanism

In this paper, an interactive evolutionary algorithm with adaptive evaluation mechanism(SVM-IEA), which based on multiclass support vector machine, is presented to solve the key problem of interactive evolutionary algorithm: the fatigue problem of user. The key to this algorithm lies in multiclass SVM using individuals evaluated by user as the training data set, then it is learned, SVM-IEA achieve the ability of displacing user to evaluate newly generated individuals. In the middle-later period of algorithm running, the user can decide whether to participate in the algorithm again until SVM-IEA obtains the optimum solutions. The proposed SVM-IEA algorithm, which can directly reduce the number of user evaluating individuals, is applied to a poster design system to empirically demonstrate its strength in alleviating user fatigue.