Re: Has this been done before or have I found something new

When you have a user interface that is used for querying in a specific domain, it is a reasonably obvious and straightforward step to eliminate any attributes and values that are not present or do not restrict the search space effectively.

A similar kind of idea appears in machine learning algorithms such as ID3. Those attributes that have the greatest power of differentiation are used first, and if after selecting some attributes, the remaining attributes that are useless are disregarded.

In fact, what you describe is very much like identifying a member using a decision tree (generated by algorithms such as ID3).

The word "algorithm" probably isn't appropriate for what you seem to be describing. It sounds more like an approach to a user interface design. You can probably find something that matches your description exactly, if you look into some expert systems, but I can't name one off the top of my head... maybe somebody else here could.

-Benjamin Johnston

"Mitchell Waite" <mitchell.dnai_at_rcn.com> wrote in message news:b0tav0$b1b$1_at_unix2.broadviewnet.net...
> Taking the chance that I am going to get flamed as a real database dummy,
a
> programmer I work with has come up with a way of quickly identifying
objects
> in a database by what he calls "smart" attribute elimination. I find it
hard
> to believe that this idea has not been done before, so I want to know if
> this is something unique or we are just repeating a basic algorithm passed
> down though the ages. So here goes.
>
> The basic idea is that a user wants to identify a particular object from a
> large set that all share common characteristics, for example he wants to
> identify a particular bird from a large group of birds in a database. The
> user selects an attribute that all the bird objects share, like eye color,
> and selects a particular value for it, say yellow. The program then
> eliminates all the bird objects that don't have that yellow value for
their
> eye color attribute, reducing the number of birds left to search. There is
> nothing earth shattering about that I but here is the twist...the program
> also eliminates all attributes that are no longer relevant to objects that
> share the selected attribute's value. For example if all birds with yellow
> eyes happen to have the same kind of beaks, the beak attribute would be
> eliminated since it will not help reduce the search any further. This the
> programmer thinks is the smart attribute elimiation process, though I find
> it hard to belive that his idea is not taught as a basic concept of
computer
> science.
>
> Taking this idea another step, say in our example that none of the birds
> remaing birds have pointed wings, but do have other kinds of wing shapes,
> the pointed wing value would be eliminated from the wing shape attribute.
> That he calls smart value emlimination. This of course is a great way to
> speed up the search but what I would like to know is this a new idea or is
> this a common algorithm taught in school or does it even have a particular
> name.
>
> Any help on clearing this up will be greatly appreciated.
>
> Sincereley,
>
> Mitchell Waite
>
>
> Publisher, Waite Group Press
> www.mitchwait.com
> mitchell.dnai_at_rcn.com
>
>
Received on Sun Jan 26 2003 - 04:03:58 CET