11 November 2013 1 10K Report

Although the term "structural representation" has been around for a long time, it appears that we don't have a good grasp of what it really means and what to expect from it. This is the main reason we are organizing workshop "What is a structural data representation? Emerging concepts of structural representation".

"We have several ambitious objectives:

‣ to address scientifically new and strategic question: how do we approach the subject of a general formalism for structural representation, i.e. which structural ‘units’ should replace the identical units of natural numbers? (e.g. structured events?)

‣ to discuss the necessary characteristics of a satisfactory candidate formalism for structural representation:

a) one of the main characteristics is the capability of the formalism to support the presently non-existent but actually the central informational concept of class (of ‘similarly’ structured objects), which is most likely defined by the generative class representation, where “generative” means “generating all class members and only them”;

b) moreover, for such class representation to be reliably learnable, the object representation must carry explicitly the generative object structure, i.e. how the agent perceives the structural generation of the object;

‣ to draw attention to the present situation: are the popular “structural” representations, indeed, structural representations?

‣ to address the applied implications of various approaches to structural representation."

http://www.cs.unb.ca/~goldfarb/SR_Workshop.pdf

More Lev Goldfarb's questions See All
Similar questions and discussions