Allow me to begin the "discussion" (we do hope that there will be one :-))
by explaining the need for it.

As some of you may know, since the emergence of AI and Pattern
Recognition, there has been, particularly during the 60's and 70's, a
widespread belief in the gradual emergence of fundamentally new formalisms
for addressing the fascinating problems arising in these areas. It appears
that, during the last two decades, this belief has, unfortunately,
materialized mainly in the introduction in the above areas of such
adjectives as "neural", "distributed", "genetic", "purposive", etc. (as
well as such nouns as "plan", "agent", etc.) without emergence of any both
radically new and more relevant formalisms.

The model under discussion--evolving transformations system (ETS)--appears
to be the first radically new formalism that addresses a number of issues
arising during the study of inductive learning processes, in particular
the issue of representation, including those of object/event
representation, class representation, and the connections between them.

In mathematics, so far, we have had "very poor" means, or formalisms,
for expressing/capturing the concept of class representation in general
and that of inductive class representation in particular
(check this statement with the known inductive learning formalisms).

ETS model was motivated, above all, by a new form of class representation
and, as a necessary consequence, also by a new form of object/event
representation. Hence the need to discuss the nature and the relevance of
the proposed formalism.

--Lev Goldfarb

Reply via email to