[Solomonov Seminar] 132. Solomonov seminar

[Marko Grobelnik]

Vabim vas na 132. Solomonov seminar, ki bo v torek, 
18. novembra 2003 ob 13:00 uri v Veliki predavalnici IJS.
Posnetki in materiali preteklih seminarjev so dostopni
na http://solomon.ijs.si

Na tokratnem seminarju bo Saso Dzeroski predstavil koncept
"induktivnih baz podatkov", ki predstavljajo poenotenje konceptov
podatkovnih baz in filozofije "data mininga". Kot realizacijo
tega koncepta bo Saso predstavil na IJS razvito resitev za
ucenje polinomskih enacb.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Saso Dzeroski, IJS:
         Inductive Queries on Polynomial Equations

Inductive databases (IDBs) contain both data and patterns.
Knowledge discovery processes can be viewed as sequences 
of inductive queries to such databases, where inductive queries 
can perform both data selection and data mining. Inductive 
queries consist of conjunctions of primitive constraints that have 
to be satisfied by target patterns. The constraints can be different 
for different pattern domains.
 
Many pattern domains, mostly concerning the discovery of frequent 
patterns, have been considered separately: these include the domains 
of frequent itemsets, episodes, Datalog queries, sequences, and 
molecular fragments. Here we consider inductive queries in the 
pattern domain of polynomial equations and present a heuristic solver 
for this domain. We first evaluate the use of this solver on standard 
regression problems, then proceed to illustrate the use of constraints 
in discovering dynamics. We finally consider IDBs containing both 
the pattern domains of equations and molecular fragments, as well 
as combining them in order to derive QSAR (Quantitative Structure-Activity Relationships) models.