Deriving the Learning Bias from Rule Properties
Jean‐Gabriel Ganascia · 1991
Abstract To guide learning, the ‘learning bias’ has to be defined with accuracy, even if it has to be modified when the results do not match expectations. This needs knowledge of the semantics of the various aspects of the learning bias: representation formalism, description language, and syntactico-semantical constraints applied to the learning assumptions. In this paper we try to show these various aspects and demonstrate how a new system Ch a r a de , provides some of these aspects with semantics, thanks to the introduction of the notion of ‘system of rules’.