Storing and Generalizing Multiple Instances While Maintaining Knowledge-Level Parallelism

One of the primary problems in knowledge representation and learning is determining how multiple instances of concepts should be organized and represented. Symbolic approaches, such as semantic networks, have been successful at representing structured knowledge for parallel access. However, such approaches have had difficulty organizing multiple instances for automatic generalization and efficient retrieval. Parallel distributed processing systems (PDP) appear to offer a solution to these problems. Unfortunately, current PDP models have not yet been able to satisfactorily represent complex knowledge structures and they remain sequential at the knowledge level. This paper presents an approach which stores multiple instances in ensembles of PDP units and organizes the ensembles in a semantic network for parallelism and structure. Thus, the best features of both styles of representation are obtained. 1 Introduction One of the central problems in knowledge representation and learning is deciding how to represent and organize multiple instances in memory. Are people able to store and retrieve every instance of an event, such as walking somewhere? Since there are enormous numbers of walking events done by a given person, or done by others and perceived by that person, retrieving every one is nearly impossible. Consequently, people must be organizing and generalizing multiple instances in some manner. 2 Previous Approaches Previous approaches to knowledge organization and generalization can be divided into two levels, shown in Figure 1. What follows is a review of symbolic and PDP approaches to the multiple instances problem. (1) Symbolic Systems: When a new event is encountered , a new token is created to represent it. For a limited number of events, storing every one does not pose a major problem. For large amounts of knowledge, similar events can be grouped together and discriminated by their differences, e.g. [Kolodner, 1984]. Unfortunately, this approach has a number of problems. First, it suffers from combinatorial explosion since there are many ways that shared features can be combined. Second, distinct tokens are needed as a final discriminant of each concept (i.e., the leaves of the tree). Third, an evaluative symbol processing mechanism is needed that decides (a) when to create a symbol, (b) how long to retain it and (c) where to index it. Finally, since each leaf is completely distinct, memory confusions do not naturally arise as a consequence of the representation. In human memory, however, confusions often do occur [Bower et a/., 1979]. (2) Parallel Distributed Processing: PDP Systems …