Evangelos Triantaphyllou and Jennifer Austin-Rodriguez
Encyclopedia of Optimization,
(P.M. Pardalos and C. Floudas, Editors), Kluwer Academic Publishers,
Boston, MA, U.S.A., Vol. 4, pp. 176-182, (2001).
Abstract:
There are many situations in which it is necessary or desirable to classify objects
into two or more mutually exclusive sets or classes. Medical diagnosis, for example,
has been the focus of numerous research efforts over the past several years. Given sets of
attributes, or relevant characteristics which describe a patient, the problem then is to extract
a pattern defined on these attributes that could be used to classify new observations. In this paper we
describe the basic steps of a Boolean functions approach in defining such patterns.
Keywords and Phrases:
Medical diagnosis, inductive inference problem, Boolean
classification problem, learning algorithms, Conjunctive Normal
Form (CNF), Disjunctive Normal Form (DNF), satisfiability (SAT)
problem, minimum number of clauses, One Clause At a Time (OCAT)
approach, positive and negative examples, GRASP approach,
randomized heuristics, missing information, unclassifiable
examples.
Index:
Medical diagnosis, breast cancer diagnosis, mammography
screening, breast tumors, inductive inference problem, Boolean
classification problem, learning algorithm, Conjunctive Normal
Form (CNF), Disjunctive Normal Form (DNF), decision trees, neural
networks, rule-based systems, satisfiability (SAT) problem,
interior point method, minimum number of DNF clauses, One Clause
At a Time (OCAT) approach, branch-and-bound, positive and
negative examples, GRASP approach, randomized heuristics,
incomplete information, missing information, unclassifiable
examples, guided learning, monotonicity, fuzziness.
(size = 1,173 KB)
