A regularization of the Frank-Wolfe method and unification of certain nonlinear programming methods (Q1338143)

The author develops a Frank-Wolfe type feasible direction algorithm for solving linearly constrained nonlinear programming problems `\(\min f(x)\)' subject to \(x\in X\), where \(X= \{x\in \mathbb{R}^n: Ax= b, x\geq 0\}\) is a nonempty polyhedron. A decomposition property and relations to other decomposition schemes are discussed. The convergence of the algorithm is demonstrated.