State-of-the-art in parallel nonlinear optimization (Q1096340)

The paper gives an excellent review on the current state-of-the-art in parallel nonlinear programming. It covers all main topics in optimization that might be suitable for parallel execution. These are: Newton's method, conjugate directions, conjugate gradients, homogeneous functions, variable metric method, multi-dimensional line searches, separable constrained optimization, decomposition of constrained problems. In each case, the basic mathematical ideas are outlined and all modifications known to the authors to take parallel calculations into account, are described briefly. In particular the availability of numerical methods, tests or comparative studies is investigated. The paper includes 66 references.