Linesearch-free algorithms for solving pseudomonotone variational inequalities
Người báo cáo: Trịnh Ngọc Hải (Đại học Bách khoa Hà Nội)
Thời gian: từ 9h sáng thứ 4 ngày 29.09.2021.
Tóm tắt: In this talk, we discuss two modified extragradient methods for variational inequalities. The first one can be applied when the Lipschitz constant of the involved operator is unknown. In contrast to the work by Hieu and Thong (J Glob Optim 70:385–399, 2018) and by Khanh (Numer Funct Anal Optim 37:1131–1143, 2016), the new algorithm does not require its step-sizes tending to zero. This feature helps to speed up our method. The second algorithm solves variational inequalities with non-Lipschitz continuous operators. Under the pseudo-monotonicity assumption, the proposed algorithm converges to a solution of the problem. In contrast to other solution methods for this class of problems, the new algorithm does not require the step sizes being square summable. Some numerical experiments show that the new algorithms are more effective than the existing ones.

Back