In this article, the authors obtain some theoretical results for 2_(IV)^(m-p) designs to have the maximum number of clear two-factor interactions by considering the number of two-factor interactions that are not clear...In this article, the authors obtain some theoretical results for 2_(IV)^(m-p) designs to have the maximum number of clear two-factor interactions by considering the number of two-factor interactions that are not clear. Several 2_(IV)^(m-p) designs with the maximum number of clear two-factor interactions, judged using these results, are provided for illustration.展开更多
频繁模式挖掘(Frequent Pattern Mining,FPM)是图数据分析、挖掘领域的核心问题之一,其目的是从大规模图数据中发现支持度不低于指定阈值的模式.传统的频繁模式挖掘算法依赖支持度进行剪枝,返回结果往往包含大量“冗余”模式;top-k模式...频繁模式挖掘(Frequent Pattern Mining,FPM)是图数据分析、挖掘领域的核心问题之一,其目的是从大规模图数据中发现支持度不低于指定阈值的模式.传统的频繁模式挖掘算法依赖支持度进行剪枝,返回结果往往包含大量“冗余”模式;top-k模式挖掘算法虽然仅返回k个频繁模式,但该类算法主要依据“客观”指标,如支持度等,对模式进行评估,难以充分反映用户的主观兴趣偏好.针对上述问题,提出一种基于主动学习的模式兴趣评估方法(Pattern Interestingness Evaluation with Active Learning,PIEAL),通过主动学习策略,从采样图上挖掘的频繁模式中选择代表性模式,并利用有限次人机交互收集用户对这些模式的偏好,进而预测模式的兴趣分数,指导算法发现用户感兴趣的模式.在人机交互环节,PIEAL采用基于成对比较的策略来收集用户对模式的偏好反馈,有效降低了用户的主观评价难度.在真实数据集上的实验结果表明,PIEAL仅需要少量的人机交互便可发现用户感兴趣的模式,其测试集准确率最高可达95%.展开更多
基金Research supported by the NNSF of China (10301015: 10571093)the SRFDP of China (20050055038)the China Portdoctoral Science Foundation (20060390169)Liu and Zhang's research was also supported by the Visiting Scholar Program at Chern Institute of Mathematics.
文摘In this article, the authors obtain some theoretical results for 2_(IV)^(m-p) designs to have the maximum number of clear two-factor interactions by considering the number of two-factor interactions that are not clear. Several 2_(IV)^(m-p) designs with the maximum number of clear two-factor interactions, judged using these results, are provided for illustration.
文摘频繁模式挖掘(Frequent Pattern Mining,FPM)是图数据分析、挖掘领域的核心问题之一,其目的是从大规模图数据中发现支持度不低于指定阈值的模式.传统的频繁模式挖掘算法依赖支持度进行剪枝,返回结果往往包含大量“冗余”模式;top-k模式挖掘算法虽然仅返回k个频繁模式,但该类算法主要依据“客观”指标,如支持度等,对模式进行评估,难以充分反映用户的主观兴趣偏好.针对上述问题,提出一种基于主动学习的模式兴趣评估方法(Pattern Interestingness Evaluation with Active Learning,PIEAL),通过主动学习策略,从采样图上挖掘的频繁模式中选择代表性模式,并利用有限次人机交互收集用户对这些模式的偏好,进而预测模式的兴趣分数,指导算法发现用户感兴趣的模式.在人机交互环节,PIEAL采用基于成对比较的策略来收集用户对模式的偏好反馈,有效降低了用户的主观评价难度.在真实数据集上的实验结果表明,PIEAL仅需要少量的人机交互便可发现用户感兴趣的模式,其测试集准确率最高可达95%.
