CAJ | 학술논문

为了寻找同时具有良好的收敛性和数值表现的共轭梯度法.将 CD 方法和 LS 方法结合,选用推广的 Wolfe 线搜索,构造出一类新的混合共轭梯度法.新的混合共轭梯度法不需要限制推广的 Wolfe 线搜索条件中的参数,但得到的下降性与 CD 法一致,具有比 CD 方法更好的收敛性,并具有全局收敛性.对新算法进行数值试验,通过与 CD 法和 LS 法的数值结果进行比较,表明新算法是可行的,尤其对大规模无约束优化问题.
위료심조동시구유량호적수렴성화수치표현적공액제도법.장 CD 방법화 LS 방법결합,선용추엄적 Wolfe 선수색,구조출일류신적혼합공액제도법.신적혼합공액제도법불수요한제추엄적 Wolfe 선수색조건중적삼수,단득도적하강성여 CD 법일치,구유비 CD 방법경호적수렴성,병구유전국수렴성.대신산법진행수치시험,통과여 CD 법화 LS 법적수치결과진행비교,표명신산법시가행적,우기대대규모무약속우화문제.
@@@@In order to find a good convergence and numerical expression of conjugate gradient method at the same time, combining the CD method and the LS method, and using general Wolfe line search, this paper has developed a new hybrid conjugate gradient method. The new hybrid conjugate gradient method is not required to restrict parameters in a general Wolfe line search, but gets the same descent property as the CD method and has a much better convergence. The new method achieves a global convergence. The numerical test of new hybrid algorithm and the comparison with numerical results of the CD method and the LS method show that the new algorithm is feasible, especially for a large-scale unconstrained optimization problem.