论文部分内容阅读
预测用户的评估特性可以有效减轻交互式进化算法中的用户疲劳问题,但基于相对尺度的用户评估制约了预测的准确性.针对这一问题.本文提出一种基于绝对尺度预测的交互式进化算法,将用户的相对评估转化成绝对评估,减少预测器学习样本中的噪声,提高预测的准确性,从而加快算法的收敛速度,更好地减轻用户疲劳.文中采用6个标准函数模拟用户.验证算法的有效性.将该算法应用于服装图像的个性化情感检索,运用符号检验方法证实采用本文所提出的算法可以获得更好的检索结果.
Predicting the user’s evaluation characteristics can effectively reduce the user fatigue in the interactive evolutionary algorithm, but the user assessment based on the relative scale restricts the accuracy of the prediction.In view of this problem, this paper proposes an interactive evolutionary algorithm based on absolute scale prediction , Which transforms the user’s relative assessment into absolute evaluation, reduces the noise in the predictor learning samples and improves the accuracy of the prediction, so as to speed up the convergence of the algorithm and reduce the user fatigue better.In this paper, six standard functions are used to simulate users. The algorithm is applied to personalized sentiment retrieval of clothing images, and the symbolic test is used to verify that the proposed algorithm can obtain better retrieval results.