|
Abstract
|
In this paper, we generalize Roos’s infeasible interior point algorithm for
linear optimization (LO) (Roos in SIAM J Optim 25(1): 102–114, 2015) to the monotone
semidefinite linear complementarity problem, based on Darvay et al.’s technique
for LO (Darvay et al. in Period Math Hung 73(1): 27–42, 2016). The symmetrization
of the search directions is based on the Nesterov–Todd scaling scheme. The algorithm
uses only one full step as feasibility step at each iteration. The derived complexity
bound coincides with the best obtained one for infeasible interior-point methods with
small updates.
|