Scholarlink

<p>In this paper, we propose an approximation Newton method for solving the low-rank matrix completion problem. We extend the traditional BFGS update technique which is used to solve the smooth nonlinear unconstrained optimization problem to matrix operation and obtain the approximate Newton direction of low-rank matrix completion problem. This technique avoids the matrix SVD decomposition, thus reducing the computational complexity of the algorithm. The backtracking line search is used such that the approximation Newton direction is computed only once and ensuring the objective function decrease monotonically. The global convergence and local superlinear convergence of the algorithm are given. The numerical results are reported and show the effectiveness of the algorithm proposed by us.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Approximation Newton method for low-rank matrix completion

  • 期刊论文

  • 发表日期：

    2025年6月6日

查看全文
  • Peng Wang,
  • Detong Zhu

摘要

In this paper, we propose an approximation Newton method for solving the low-rank matrix completion problem. We extend the traditional BFGS update technique which is used to solve the smooth nonlinear unconstrained optimization problem to matrix operation and obtain the approximate Newton direction of low-rank matrix completion problem. This technique avoids the matrix SVD decomposition, thus reducing the computational complexity of the algorithm. The backtracking line search is used such that the approximation Newton direction is computed only once and ensuring the objective function decrease monotonically. The global convergence and local superlinear convergence of the algorithm are given. The numerical results are reported and show the effectiveness of the algorithm proposed by us.