Ephemeral Resource Constraints in Optimization

摘要: Constraints in optimization come traditionally two types familiar to most readers: hard and soft. Hard constraints delineate absolutely between feasible infeasible solutions, whereas soft essentially specifyadditional objectives. In this chapter, we describe a third type of constraint, much less only investigated recently, which call ephemeral resource (ERCs). ERCs differ from the other three major ways. (i) The are dynamic or temporary (i.e., may be active not active), occur during optimization—they do affect feasibility final solutions. (ii) Solutions violating cannot evaluated on objective function—in fact that is their main defining property. (iii) usually function previous solutions evaluated, bringing time-linkage aspect optimization. We explain with examples how these arise real-world problems, especially when solution evaluation depends experimental processes (i.e. “closed-loop optimization”). Using theoretical model based Markov chains, effects evolutionary search, e.g., drift search direction, described. Next, number strategies for coping summarized, evidence robustness provided. section, look future consider many open questions there new area.