A Long Step Primal-Dual Path Following Method for Semidefinite Programming
In this paper, we consider an extension of the long step primal-dual
path followin g method developed by Jansen et al. for linear
programming to semidefinite programming. By extending the proximity
measure and using the primal-dual logarithmic function as the merit
function, we present an algorithm which finds an $\varepsilon$-optimal
solut ion in at most $O(n|\ln \varepsilon |)$ iterations.
Science Report 96009, Dept. of Applied Math., Tsinghua Univ.,
Beijing 100084, Chin a. March, 1996.