论文部分内容阅读
稀疏logistic回归,是机器学习中一类重要的问题,它在控制论、管理科学和互联网等领域有着广泛的应用.本文提出了一种改进的基于分象限学习的拟Newton算法来求解稀疏logistic回归问题.新算法采用著名的Barzilai-Borwein步长策略自适应地近似代替目标函数的Hesse阵,并利用目标函数的整体梯度信息来构造拟Newton向量.在适当的条件下,证明了新算法的全局收敛性.数值实验表明新算法是可行的,并且是有效的.
Sparse logistic regression is a kind of important problem in machine learning and it has a wide range of applications in cybernetics, management science and the Internet.In this paper, we propose an improved Quasi-Newton algorithm based on quadrant quadrant learning to solve sparse logistic regression Problem.The new algorithm uses the well-known Barzilai-Borwein step strategy to adaptively approximate the Hesse matrix of the objective function and constructs the quasi-Newton vector using the overall gradient information of the objective function. Under appropriate conditions, Convergence. Numerical experiments show that the new algorithm is feasible and effective.