本研究基于KISS(keep it simple and stupid)算法,利用似然比测试直接为矩阵模式定义度量,解决了现有大多数度量学习算法需要经过复杂优化过程的问题。通过在似然比测试中有目的地引入矩阵正态分布,该度量无需将矩阵模式通过向量化的方...本研究基于KISS(keep it simple and stupid)算法,利用似然比测试直接为矩阵模式定义度量,解决了现有大多数度量学习算法需要经过复杂优化过程的问题。通过在似然比测试中有目的地引入矩阵正态分布,该度量无需将矩阵模式通过向量化的方法变成向量模式,因而具有如下优点:(1)能够避免维数灾难;(2)比KISS更鲁棒;(3)无需计算大矩阵的逆和特征值分解,因此计算远快于KISS算法。最终的实验验证了该算法的优势。展开更多
In this paper, the authors generalize the definition of χ 2 distribution and introduce a quasi χ 2 distribution, and then prove several properties of it, find the necessary and sufficient conditions of i...In this paper, the authors generalize the definition of χ 2 distribution and introduce a quasi χ 2 distribution, and then prove several properties of it, find the necessary and sufficient conditions of independence about multivariate normal distributions, matrix normal distributions and two parts of the Wishart distribution. 展开更多
文摘本研究基于KISS(keep it simple and stupid)算法,利用似然比测试直接为矩阵模式定义度量,解决了现有大多数度量学习算法需要经过复杂优化过程的问题。通过在似然比测试中有目的地引入矩阵正态分布,该度量无需将矩阵模式通过向量化的方法变成向量模式,因而具有如下优点:(1)能够避免维数灾难;(2)比KISS更鲁棒;(3)无需计算大矩阵的逆和特征值分解,因此计算远快于KISS算法。最终的实验验证了该算法的优势。
文摘In this paper, the authors generalize the definition of χ 2 distribution and introduce a quasi χ 2 distribution, and then prove several properties of it, find the necessary and sufficient conditions of independence about multivariate normal distributions, matrix normal distributions and two parts of the Wishart distribution.
