A novel adaptive detection scheme both for point-like and distributed targets in the presence of Gaussian disturbance in the partial y homogeneous environment （PHE） is proposed. The novel detection scheme is based o...A novel adaptive detection scheme both for point-like and distributed targets in the presence of Gaussian disturbance in the partial y homogeneous environment （PHE） is proposed. The novel detection scheme is based on the orthogonal projection technique. Both the case of known covariance matrix structure and the case of unknown covariance matrix structure are con-sidered. For the former case, the closed-form statistical pro-perty of the novel detectors is derived. When the covariance matrix is unknown, the corresponding detectors have higher probabilities of detection （PDs） than their natural competitors. Moreover, they ensure constant false alarm rate （CFAR） property.展开更多
时间序列早期分类(ETSC)有两个矛盾的目标:早期性和准确率。分类早期性的实现,总是以牺牲它的准确率为代价。现有基于优化的多变量时间序列(MTS)早期分类方法,虽然在成本函数中考虑了错误分类成本和延迟决策成本,却忽视了MTS数据集样本...时间序列早期分类(ETSC)有两个矛盾的目标:早期性和准确率。分类早期性的实现,总是以牺牲它的准确率为代价。现有基于优化的多变量时间序列(MTS)早期分类方法,虽然在成本函数中考虑了错误分类成本和延迟决策成本,却忽视了MTS数据集样本之间的局部结构对分类性能的影响。针对这个问题,提出一种基于正交局部保持映射(OLPP)和成本优化的MTS早期分类模型(OLPPMOAE)。首先,使用OLPP将MTS样本前缀映射到低维空间,保持原数据集的局部结构;其次,在低维空间训练一组高斯过程(GP)分类器,生成训练集每个时刻的类概率;最后,使用粒子群优化(PSO)算法从这些类概率中学习停止规则中的最优参数。在6个MTS数据集上的实验结果表明,在早期性基本持平的情况下,OLPPMOAE的准确率显著高于基于成本的R1_C_(lr)(stopping Rule and Cost function with regularization term l_(1)and l_(2))模型,平均准确率能够提升11.33%~15.35%,调和均值(HM)能够提升4.71%~9.01%。因此,所提模型能够以较高的准确率尽早地分类MTS。展开更多
目的探讨orthogonal projection to latent structures(OPLS)方法的原理、特点及其在代谢组学高维数据分析中的应用。方法通过R语言编程实现OPLS方法,利用模拟试验探索OPLS的特性及适用条件,并通过实际数据进行验证。结果利用一个OPLS...目的探讨orthogonal projection to latent structures(OPLS)方法的原理、特点及其在代谢组学高维数据分析中的应用。方法通过R语言编程实现OPLS方法,利用模拟试验探索OPLS的特性及适用条件,并通过实际数据进行验证。结果利用一个OPLS预测主成分的模型拟合效果与利用偏最小二乘(PLS)多个主成分的模型拟合效果相同,同时具有较好的判别能力,其得分图的可视化效果优于PLS。结论 OPLS能够有效去除自变量矩阵X中与因变量Y无关的信息,使模型变得简单、易于解释,同时具有较好的可视化效果,可有效地用于代谢组学数据分析中。展开更多
基金supported by the National Natural Science Foundation of China(61102169)the Outstanding Youth Fund of the National Natural Science Foundation of China(60925005)
文摘A novel adaptive detection scheme both for point-like and distributed targets in the presence of Gaussian disturbance in the partial y homogeneous environment （PHE） is proposed. The novel detection scheme is based on the orthogonal projection technique. Both the case of known covariance matrix structure and the case of unknown covariance matrix structure are con-sidered. For the former case, the closed-form statistical pro-perty of the novel detectors is derived. When the covariance matrix is unknown, the corresponding detectors have higher probabilities of detection （PDs） than their natural competitors. Moreover, they ensure constant false alarm rate （CFAR） property.
文摘时间序列早期分类(ETSC)有两个矛盾的目标:早期性和准确率。分类早期性的实现,总是以牺牲它的准确率为代价。现有基于优化的多变量时间序列(MTS)早期分类方法,虽然在成本函数中考虑了错误分类成本和延迟决策成本,却忽视了MTS数据集样本之间的局部结构对分类性能的影响。针对这个问题,提出一种基于正交局部保持映射(OLPP)和成本优化的MTS早期分类模型(OLPPMOAE)。首先,使用OLPP将MTS样本前缀映射到低维空间,保持原数据集的局部结构;其次,在低维空间训练一组高斯过程(GP)分类器,生成训练集每个时刻的类概率;最后,使用粒子群优化(PSO)算法从这些类概率中学习停止规则中的最优参数。在6个MTS数据集上的实验结果表明,在早期性基本持平的情况下,OLPPMOAE的准确率显著高于基于成本的R1_C_(lr)(stopping Rule and Cost function with regularization term l_(1)and l_(2))模型,平均准确率能够提升11.33%~15.35%,调和均值(HM)能够提升4.71%~9.01%。因此,所提模型能够以较高的准确率尽早地分类MTS。
