Design and Management of Complex Technical Processes and Systems by means of Computational Intelligence Methods Predicting the Solution Quality in Noisy Optimization
نویسندگان
چکیده
Noise often occurs in real-world optimization tasks. Although evolution strategies and other evolutionary algorithms are thought to be robust against the effects of noise, even their performance is degraded. Generally, one observes a reduction of the convergence velocity and a deterioration of the final solution quality. There are two local progress measures describing the performance of evolution strategies from one generation to the next. The first – the progress rate – is defined directly on the object parameter space, whereas the second, called the quality gain, operates in the space of the fitness values instead. Although both are local performance measures, they can be utilized to derive evolution criteria and steady state conditions. The latter can be used to predict the final solution quality. We will determine the quality gain for (1, λ)-ES in the presence of noise. After the derivation of the steady state conditions, we will develop an equation describing the final fitness error for some test functions. The results will be extended to (μ/μI , λ)-ES and compared with ES runs using σ-self-adaptation as well as CSA.
منابع مشابه
Design and Management of Complex Technical Processes and Systems by means of Computational Intelligence Methods Lower bounds for hit-and-run optimization
and was printed with financial support of the Deutsche Forschungsgemeinschaft.
متن کاملDesign and Management of Complex Technical Processes and Systems by means of Computational Intelligence Methods Evolutionary Algorithms and Matroid Optimization Problems
We analyze the performan e of evolutionary algorithms on various matroid optimization problems that en ompass a vast number of e iently solvable as well as NP-hard ombinatorial optimization problems (in luding many well-known examples su h as minimum spanning tree and maximum bipartite mat hing). We obtain very promising bounds on the expe ted running time and quality of the omputed solution. O...
متن کاملDesign and Management of Complex Technical Processes and Systems by means of Computational Intelligence Methods A Natural and Simple Function Which is Hard For All Evolutionary Algorithms
Evolutionary algorithms (EAs) are randomized search strategies which have turned out to be efficient for optimization problems of quite different kind. In order to understand the behavior of EAs, one also is interested in examples where EAs need exponential time to find an optimal solution. Until now only artificial examples of this kind were known. Here an example with a clear and simple struc...
متن کاملDesign and Management of Complex Technical Processes and Systems by means of Computational Intelligence Methods Computing Minimum Cuts by Randomized Search Heuristics
We study the minimum s-t-cut problem in graphs with costs on the edges in the context of evolutionary algorithms. Minimum cut problems belong to the class of basic network optimization problems that occur as crucial subproblems in many real-world optimization problems and have a variety of applications in several different areas. We prove that there exist instances of the minimum s-t-cut proble...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2004