Integration of Expert’s Preferences in Pareto Optimization by Desirability Function Techniques
نویسندگان
چکیده
Many real-world problems have a multiobjective character. A-posteriori techniques such as multiobjective evolutionary algorithms (MOEA) generate best compromise solution sets, i.e. Pareto-fronts and Pareto-sets. Classic MOEA are able to find quite efficiently good approximations of the complete Pareto-fronts also for very complex problems. In real-world applications only small sections of the complete front are of practical interest. Desirability functions are a flexible, intuitive and mathematically sound technique to focus on Pareto-front segments without loosing the power of the a-posteriori population search methods. A systematic analysis of typical multiobjective evolutionary algorithms applied to a set of test functions and a real-world problem, the mold temperature control design problem, using the desirability function technique is presented.
منابع مشابه
Desirability factors of work desk arrangement from the viewpoint of employees by the analysis of space syntax indices
Using the results of the assessment of desirability from the viewpoint of employees and the computer analysis of space syntax, the desirability factors of work desk arrangement in office space have been analyzed in the present study. Firstly a pictorial questionnaire was distributed among 113 employees in two offices of Tabriz city to obtain the desirable sitting places in the room (considering...
متن کاملPreferences and their application in evolutionary multiobjective optimization
The paper describes a new preference method and its use in multiobjective optimisation. These preferences are developed with a goal to reduce the cognitive overload associated with the relative importance of a certain criterion within a multiobjective design environment involving large numbers of objectives. Their successful integration with several genetic algorithm–based design search and opt...
متن کاملOptimization of the strain distribution in the roll forming process using the desirability function and finite element methods
Defects of the roll forming process are affected by amount and situation of the strains distribution. The effect of the process parameters on the strain distribution in the round cross section roll forming process has been studied. Finite element and response surface method have been used for process modelling. Then desirability functions approach and overlaid counter plots have been employed f...
متن کاملPreference-based Reinforcement Learning
This paper investigates the problem of policy search based on the only expert’s preferences. Whereas reinforcement learning classically relies on a reward function, or exploits the expert’s demonstrations, preference-based policy learning (PPL) iteratively builds and optimizes a policy return estimate as follows: The learning agent demonstrates a few policies, is informed of the expert’s prefer...
متن کاملParallel Universes: Multi-Criteria Optimization
In this paper parallel universes are defined by their relation to multi-criteria optimization combined with an explicit or implicit link for the unambiguous identification of an optimum. As an explicit link function the desirability index is introduced. Desirabilities are also used for restricting the Pareto set to desired parts.
متن کامل