A New Hybrid Algorithm of Trust Region Methods

2014 ◽  
Vol 680 ◽  
pp. 442-446
Author(s):  
De Qing Zhang ◽  
Pei Pei Zhou ◽  
Qing Hua Zhou
Keyword(s):  
Numerical Experiments ◽  
Trust Region ◽  
Acute Angle ◽  
Step Size ◽  
Interaction Point ◽  
Trust Region Algorithms ◽  
Value Range ◽  
One Step ◽  
Current Interaction ◽  
Negative Gradient

In the solution of trust region subproblem within the classical trust region algorithms, the centre of sphere is the current interaction point and one step-size is the upper bound. Considering that only with the negative gradient direction to acute angle may reduce the function value, we introduce the parameter to control of the centre of sphere and the radius. Based on the numerical experiments, obtains the value range of the parameter. The numerical evaluation demonstrates the validity of the new trust region algorithms.

scholarly journals A randomized adaptive trust region line search method

2020 ◽  
Vol 10 (2) ◽  
pp. 259-263
Author(s):  
Saman Babaie-Kafaki ◽  
Saeed Rezaee
Keyword(s):  
Simulated Annealing ◽  
Numerical Experiments ◽  
Line Search ◽  
Optimization Problems ◽  
Trust Region ◽  
Search Method ◽  
Test Problems ◽  
Unconstrained Optimization Problems ◽  
Line Search Method ◽  
Region Line

Hybridizing the trust region, line search and simulated annealing methods, we develop a heuristic algorithm for solving unconstrained optimization problems. We make some numerical experiments on a set of CUTEr test problems to investigate efficiency of the suggested algorithm. The results show that the algorithm is practically promising.

THE FORWARD MAGNETOTELLURIC PROBLEM FOR AN INHOMOGENEOUS AND ANISOTROPIC STRUCTURE

Geophysics ◽  
1974 ◽  
Vol 39 (1) ◽  
pp. 56-68 ◽  
Author(s):  
Flavian Abramovici
Keyword(s):  
Three Dimensional ◽  
Middle Layer ◽  
Conductivity Tensor ◽  
Dimensional Structure ◽  
Impedance Tensor ◽  
Variable Step Size ◽  
Step Size ◽  
Variable Step ◽  
One Step ◽  
Numerical Integration Procedure

The impedance tensor corresponding to the magnetotelluric field for a nonisotropic one‐dimensional structure is given in terms of the solutions of a sixth‐order differential system. The conductivity tensor is three‐dimensional. Its components depend upon depth only in an arbitrary manner such that the corresponding matrix is positive definite. The impedance tensor components are found by a numerical integration procedure based on a set of one‐step methods and a variable step‐size to insure a given accuracy in the final result. Calculations were made for three models having sharp boundaries and also transitional layers. The first of these models has a middle layer of high conductivity, sandwiched between two layers of linearly varying conductivity, while in the second model the middle layer has a very low conductivity. In the third model the conductivity tensor is three‐dimensional and is linearly varying in one of the layers.

Global convergence of trust-region algorithms for convex constrained minimization without derivatives

2013 ◽  
Vol 220 ◽  
pp. 324-330 ◽  
Author(s):  
P.D. Conejo ◽  
E.W. Karas ◽  
L.G. Pedroso ◽  
A.A. Ribeiro ◽  
M. Sachine
Keyword(s):  
Global Convergence ◽  
Trust Region ◽  
Constrained Minimization ◽  
Trust Region Algorithms
Sign in / Sign up

Export Citation Format

Share Document