Symbol Processing Systems, Connectionist Networks, and Generalized Connectionist Networks

摘要: Many authors have suggested that SP (symbol processing) and CN (connectionist network) models offer radically, or even fundamentally, different paradigms for modeling intelligent behavior (see Schneider, 1987) the design of systems. Others argued little to contribute our efforts understand intelligence (Fodor & Pylyshyn, 1988). A critical examination popular characterizations suggests neither these extreme positions is justified. There are many advantages be gained by a synthesis best both approaches in The Generalized connectionist networks (GCN) (alternately called generalized neuromorphic systems (GNS)) introduced this paper provide framework such synthesis. 1. Symbol Processing Architectures Intelligent Systems symbol processing approach was summarized Newell (1980) Simon (1972) terms what they physical Fodor (1976) he language thought. In framework, perception cognition tantamount acquiring manipulating symbolic representations. It structures constitute representations counterpart structure brain and/or brain’s internal states. Models developed within typically (but do not be) based on von Neumann serial stored program model computation. Popular interpretations definition often overly restrictive, appear exclude (for no good reason) perceive, learn, reason with non-symbolic (e.g., iconic analogic) representations, using numerically-encoded probabilistic fuzzy inference structures. following sections critically examine conceptions major strengths weaknesses hhhhhhhhhhhhhhhhhhhhhhhhhhhhh This draft currently under review. Comments suggestions improvement will appreciated. work partially supported University Wisconsin-Madison Graduate School Iowa State College Liberal Arts Sciences.