[Coin-discuss] Multicriteria optimization

[Lou Hafer]

Emilia,

	Well, my reaction, up until the last paragraph, was ``looking forward
to details so we can provide some reaction.'' But I'll try a few thoughts
and questions.

	It's true that COIN leans toward mathematical models -- that's the
primary expertise of the active developers to date.  As you say, there are other
methods that require different (not necessarily less rigorous) models.  One area
that would certainly be of interest is constraint logic programming, and the
intersection of CLP and mathematical programmng.  And one can point to many
different state space search methods (tabu, probabilistic hill climbing) which
require little beyond the ability to make a change and evaluate the effect.

	I'm unsure what you have in mind by `imperative model'.  Could be I'm
betrayed by my computer science upbringing.  `Imperative' to me means that the
user states explicitly the changes they want to apply to an object (in effect,
states an algorithm to achieve the goal).  As opposed to `declarative', where
one states the qualities the object must possess and leaves it to the underlying
solution engine to come up with the object.  In this sense, CLP is declarative,
as is a linear or nonlinear mathematical model.

	So I'm left with the question, ``What do you have in mind when you
suggest an imperative interface for multicriteria optimisation?'' I can envision
an interactive interface with access to a selection of underlying algorithms. 
And I can envision a Pandora's box of representational issues.

	Can you give an example of what you have in mind, or a pointer to
something with more detail?

						Lou