Optimization with a Correct Convergence Method
Our test results proved that non-target convergence methods have a premature convergence problem. And if it is true to NM simplex optimization, then it may be true to other search methodologies, such as Monte Carlo, Simulated Annealing, and Particle Swarm Optimization, too.
Target convergence method avoids the premature convergence problem, but it is only applicable to functions with a given minimum, without local minima, without flat response areas. The best application examples are the 2-dimensional McKinnon functions and many other standard test functions. The target convergence method together with the termN method may be useful for parameter optimization, where the target minimum can be set to zero. The parameter optimization will be terminated by a user-specified maximum step number when the target minimum is not zero. Otherwise, the target convergence method will converge. The converge should be more efficient than..
 |
Dehua Joined: Aug-20-2014 |
Dehua Wang is a physical chemist. He received his Bachelor of Science degree from the University of Science and Technology of China in 1963 and received his Ph.D. in physical chemistry from Syracuse University. Syracuse, New York in 1985. Dehua is presently a senior scientist at California Sharon Company. For more information, see Marquis Who’sWho in America, 52nd Edition, New Providence, NJ, USA, p4505, 1998.
Xiaolong Xu is a chemist. She received her Bachelor of Science degree from the University of Science and Technology of China in 1963. She worked at Emory University (Atlanta, Georgia) for more than ten years. She is presently a manager at California Sharon Company, working in sales and finance.
