POSSIBILISTIC CONDITIONAL-INDEPENDENCE - A SIMILARITY-BASED MEASURE AND ITS APPLICATION TO CAUSAL NETWORK LEARNING

A definition for similarity between possibility distributions is intro duced and discussed as a basis for detecting dependence between variab les by measuring the similarity degree of their respective distributio ns. This definition is used to detect conditional independence relatio ns in possibility distributions derived from data. This is the basis f or a new hybrid algorithm for recovering possibilistic causal networks . The algorithm POSS-CAUSE is presented and its applications discussed and compared with analogous developments in possibilistic and probabi listic causal networks learning. (C) 1998 Elsevier Science Inc.