Lazyweb: can anyone tell me about symmetric classification


Classification involves dividing a set of items into classes based on their attributes. I've been trying to apply this to modelling autism, without much success. I now suspect that part of the reason for my failure is that the traditional assumptions in classification problems do not often apply in the real world.

Traditional (computer aided) classification (see eg assumes a large number of items and a small number of attributes. But consider:

The traditional model of classification breaks down in the following ways:

Furthermore, these problems are often symmetric: The problem is just as meaningful if you call the attributes items and the items attributes.

I'm interested in classification strategies that treat the problem symmetricaly. Ring any bells for you?


Lee points me to this Clay Shirky essay, in which Shirky points out the flaws of hierarchical classification and advocates tagging as an alternative.

Jiri pointed me to one of his blog entries which touches on similar concerns, and suggests the keyword "bigraph".

pfh notes that Singular Value Decomposition is suitably symmetric. I know this has been applied to at least the words-and-documents problem.