Demonstartion of GAs/RS3

Over the years, genetic algorithms (GAs) have proven useful in varieties of search and optimization problems. There are many theoretical and empirical studies that investigate or present ways to improve the performance of conventional GAs for difficult function optimization problems such as those posed by multimodal and deceptive functions.
If a solution obtained by some search technique is very sensitive to small perturbations of its parameter values, it may not be good to use this solution in certain situations. For example, consider the problem of designing the optimal parameter values of a process control plant. Suppose, by some technique, we determine the parameter values which yield very high performance from the plant. Now if the activity of the plant changes heavily due to small variation of the parameter values, then it is very risky to use such a parameter set for the plant, because in practice some noise will always be involved with the parameter values. Let us consider another case. If the performance of a product is highly sensitive to the precision of its parts, then the product will be very difficult or costly to produce by machines because each and every machine has limited capability to handle precision. Thus in many optimization tasks, there is a need to determine solutions whose value will not change much due to small variation of the parameter values.
We describe this type of solution as robust. This study provides the basic concept of a new technique which extends the application of GAs to domains that require identification of robust solutions. We call this new technique GAs/RS3: GAs with a robust solution searching scheme.
In the GAs/RS3 perturbations (noise) are added to the phenotypic parameter values while evaluating the functional value or fitness of individuals. Simulation results on various functions having broad and sharp peaks show the potential of the proposed concept.
Please see papers in the publication list page by downloading ps or pdf files.
Usage

Press the button below. You can choose three test functions from pull down menu. Please compare GA/RS3 with Simple GA. You can choose GA model from the pull down menu.