Fundamentals of Statistics contains material of various lectures and courses of H. Lohninger on statistics, data analysis and chemometrics......click here for more.


Optimization - Visualization of the Response Function

Response functions are often very complex and depend on several variables, which prevents their visualization. In fact, an optimization problem could quickly be solved if the response function could be plotted.
 

It is important to keep in mind that the researcher does not usually know much about the response surface of the search problem. One could compare this situation to finding the peak of a particular landscape in a dense fog. This fog will prevent the climber from seeing the whole landscape, so she has to decide from that which can be observed in the vicinity where to go next in order to find the maximum. This may well result in a situation where the climber may reach a top which is not the highest peak. This is a very important point and one of the major problems in optimization tasks. Very often (especially in complex systems) it is even impossible to detect whether one reached the global optimum or just a local one.