【摘要】 通过将逻辑斯蒂回归转化为非线性方程组,本文设计了求解逻辑斯蒂回归的惯性PRP-FR混合无导数投影法.该方法采用惯性技巧、PRP-FR混合策略和超平面投影技巧来加速收敛速度,同时不需要计算逻辑斯蒂回归函数的Hessian阵,因此适合求解大规模逻辑斯蒂回归.分析了方法的全局收敛性与收敛速度,最后通过Heart Disease数据集和LIBSVM数据集验证了方法的有效性.更多还原

【Abstract】 By transforming logistic regression into a system of nonlinear equations, this paper designs an inertia PRP-FR hybrid derivative-free projection method for logistic regression. This method uses inertia techniques, PRP-FR hybrid strategies, and hyperplane projection techniques to accelerate convergence speed, while not requiring the calculation of the Hessian matrix of the logistic regression function,making it suitable for solving large-scale logistic regression. The global convergence and convergence rate of the method are also analyzed, and its effectiveness is finally verified through the Heart Disease dataset and the LIBSVM dataset.更多还原