Journal of Xidian University ›› 2024, Vol. 51 ›› Issue (3): 147-157.doi: 10.19665/j.issn1001-2400.20240301

• Computer Science and Technology & Artificial Intelligence • Previous Articles     Next Articles

Relatively accelerated stochastic gradient algorithm for a class of non-smooth convex optimization problem

ZHANG Wenjuan1(), FENG Xiangchu(), XIAO Feng3(), HUANG Shujuan3(), LI Huan1()   

  1. 1. School of Sciences,Xi’an Technological University,Xi’an 710021,China
    2. School of Mathematics and Statistics,Xidian University,Xi’an 710071,China
    3. School of Computer Science and Engineering,Xi’an Technological University,Xi’an 710021,China
  • Received:2023-01-03 Online:2024-03-26 Published:2024-03-26
  • Contact: XIAO Feng E-mail:zhangwenjuan@xatu.edu.cn;xcfeng@xidian.edu.cn;544070146@mail.xidian.edu.cn;349242386@qq.com;1186858891@qq.com

Abstract:

The first order method is widely used in the fields such as machine learning,big data science,computer vision,etc.A crucial and standard assumption for almost all first order methods is that the gradient of the objective function has to be globally Lipschitz continuous,which,however,can’t be satisfied by a lot of practical problems.By introducing stochasticity and acceleration to the vanilla GD (Gradient Descent) algorithm,a RASGD (Relatively Accelerated Stochastic Gradient Descent) algorithm is developed,and a wild relatively smooth condition rather than the gradient Lipschitz is needed to be satisfied by the objective function.The convergence of the RASGD is related to the UTSE (Uniformly Triangle Scaling Exponent).To avoid the cost of tuning this parameter,a ARASGD(Adaptively Relatively Accelerated Stochastic Gradient Descent)algorithm is further proposed.The theoretical convergence analysis shows that the objective function values of the iterates converge to the optimal value.Numerical experiments are conducted on the Poisson inverse problem and the minimization problem with the operator norm of Hessian of the objective function growing as a polynomial in variable norm,and the results show that the convergence performance of the ARASGD method and RASGD method is better than that of the RSGD method.

Key words: convex optimization, nonsmooth optimization, relatively smooth, stochastic programming, gradient method, accelerated stochastic gradient descent

CLC Number: 

  • O24

Baidu
map