No label defined (Q3325435)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3856799
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 3856799

      Statements

      0 references
      1983
      0 references
      survey
      0 references
      multi-objective methods
      0 references
      convergence results
      0 references
      [For the entire collection see Zbl 0528.00011.] This paper describes nine multi-objective methods based on a combination of mathematical programming and interactions with the decision maker. The methods use information acquired from the decision maker on preferences between pairs of actions presented to a decision maker up to a given stage to produce surrogate optimization or efficient set problems, which are then used further to explore the action space, with the procedures repeated until optimal or acceptable solutions are obtained. The method of Dyer, Feinberg and Geoffrion assumes a convex compact action space, a concave and continuously differentiable value function, unknown explicitly, but governing the decision maker's responses. The surrogate optimization problem is the linear approximation at the current point. The method produces a sequence of improving actions, and limit points of sequences are optimal. The method of the author assumes a compact action space and continuous objective functions, with a linear value function, unknown explicitly, but governing the decision maker's responses. At each stage the surrogate optimization problem selects the next action to include, and the process stops at an optimal solution when the optimization index is zero. The information gained on preferences at each stage is used to determine the region of weighting factors consistent with preferences to date. Certain convergence results are presented. The remaining seven methods are those of Wallenius and Zionts, (with assumptions as for the author's method, but with the feasible action space being a polytope, and linear objective functions, and uses a vertex search procedure), Steuer (with assumptions as for Wallenius and Zionts, but with initial bounds on the weight factors, and with the conversion of the initial optimization problem into an efficient solution search problem), Gabbani and Magazine (with the same assumptions as for Wallenius and Zionts, but with integer restrictions, and using a simplicial generation procedure to restrict the feasible region of weighting factors), Morris and Oren (with polyhedral objective function space, continuous objective functions, and a quasi concave value function, and which allows the decision maker to move outside the feasible action space to provide useful information), Marcotte and Soland (which deals with finite action spaces, and introduces a type of branch and bound procedure), Stewart (which deals with finite action spaces, and a class of utility functions, with unknown parameter, subject to Bayesian updating on its likelihood function), and finally, Yu (who uses domination cone ideas, with the decision maker having an unknown, but implicit, domination cone governing his partial ordering of actions).
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references