In the paper we discuss a new approach for learning classification rules from examples. We sketch out the algorithm JoJo and its extension to a four step procedure which can be used to incrementally refine a set of classification rules. Incorrect rules are refined, the entire rule set is completed, redundant rules are deleted and the rule set can be minimized. The first two steps are done by ap...

Bilevel decision addresses the problem in which two levels of decision makers, each tries to optimize their individual objectives under constraints, act and react in an uncooperative, sequential manner. Such a bilevel optimization structure appears naturally in many aspects of planning, management and policy making. However, bilevel decision making may involve many uncertain factors in a real w...

Ensembles are a set of classification models that, when combined, produce better predictions than when used by themselves. This chapter proposes a new evolutionary algorithm-based method for creating an ensemble of rule sets consisting of two stages. First, an evolutionary algorithm (more precisely, a genetic programming algorithm) is used to automatically create complete rule induction algorit...

We present an approach to learning sets of recursive rules based on analytical inductive programming. We propose that our approach can be used within cognitive architectures to model regularity detection and generalization over experience. Induced rule sets represent the knowledge underlying systematic behavior in complex situations. Such rule sets can explain systematicity and productivity of ...

We present a technique for evaluating classifications by geometric comparison of rule sets, Rules ~e represented as objects in an n-dimensional hyperspace. The similarity of classes is computed from the overlap of the geometric class descriptions, The system produces a correlation matrix that ̄ indicates the degree of similarity between each pair of classes. The technique can be applied to class...

Rules are commonly used for classification because they are modular, intelligible and easy to learn. Existing work in classification rule learning assumes the goal is to produce categorical classifications to maximize classification accuracy. Recent work in machine learning has pointed out the limitations of classification accuracy: when class distributions are skewed, or error costs are unequa...

This derivation of (2) is not valid, because this duality applies to inferences and theorems only, not to definitions. In fact, (1) should be treated as an axiom. Accordingly, the duality of Lemma 3.3.1 would not be valid until the dual (2) of (1) is proved independently, as we do below. Our new proof will make use of Proposition 3.3.3, the original proof of which is based on duality. So, we mu...

Answering a recent question posed by Gregori, Morillas and Sapena (“On a class of completable fuzzy metric spaces”, Fuzzy Sets and Systems, 161 (2010), 2193–2205) we present two examples of non strong fuzzy metrics (in the sense of George and Veeramani).

Verb modality presents a major processing obstacle in any NLP application, and can be overcome either by exhaustively listing all possible modal variants of any given verb, or by deriving any modality-based valency alterations in situ. Given the combinatorial complexity of any exhaustive listing attempt, we adopt this second methodology in proposing a modular set of valency frame transformation...