Parallel Distributed Processing : Bridging the Gap between Human and Machine Intelligence

概要

論文の詳細を見る
The Parallel Distributed Processing approach to modeling intelligent information processing grew out of the feeling that conventional computational approaches to modeling intelligent information processing were not well suited to capturing several key aspects of human information processing abilities. In this paper, we describe these key aspects of human processing abilities, and we note how they can be captured in parallel-distributed processing models (otherwise known as connectionist models or neural network models). A key feature of these models is their ability to make effective use of graded or continuous-valued information. We then note how the ability to make use of graded information is exploited in several current connectionist AI research projects. The final section of the paper illustrates how machines that make use of graded information may increase our taxonomy of basic computational machine types. We introduce the notion of the graded state machine and argue that it has characteristics which place its capabilities between finite-state and recursive computational devices.
社団法人人工知能学会の論文
1990-01-01