Optimization does not work, what can be done?
The application of optimization methods does not really work well in my field. What can be done to improve the situation?
If the application of optimization methods does not work well, most of the time the proper formulation of the target functions and optimization goals is the number one source of errors and improvements.
Herefore the target functions must include all relevant rating criteria and be able to balance the potential conflicts in the criteria in a way that the resulting target function correlates best possibly with the expectation of the executing engineers and customers. This is a self-contained regression task, which often needs at least the same attention and passion than the main optimization problem. Here again Machine Learning approches may be the methods of choice. These are extremely helpful for the identification and judgement of the right performance criteria for the formulation of a proper optimization target functions according the engineer's, expert's and customer's rating. Machine Learning methods can e.g. easily and efficiently be trained to reflect the experts rating and then be used directly within the automated optimization procedures.
Last update on 2016-12-04 by Andreas Kuhn.