Högskolan i Skövde

This paper outlines the ongoing work in Reactive Object-oriented Database Systems between the Departments of Computer Science in the University of Exeter(UK) and the University of Skovde(Sweden). The group is currently designing a monitoring system based on a reactive object oriented database with the objective of supporting efficient interaction between the active DBMS and applications (including intelligent systems). Initial work has centered on a prototype reactive object-oriented system built on top of ONTOS, a commercial OODBMS which has C++ as its base language. The prototype is referred to as ACOOD (ACtive Object Oriented Database system). We briefly discuss this prototype, showing how reactive behaviour has been incorporated into a full OODBMS albeit with some restrictions. We also outline our plans for its future extensions, and how these are motivated.

A connectionist architecture is outlined which makes use of RAAM to generate representations for objects in inheritance networks and extended learning to make such representations context-sensitive. The architecture embodies inheritance quite differently by relying on associative similarities and regions in representational space. The model avoids many of the problems identified for traditional inheritance.

Nonmonotonic reasoning is a core problem in AI. An example of nonmonotonic reasoning is the type of default reasoning which occurs with inheritance structures which allow exceptions. This paper describes a connectionist model of a hierarchical inheritance structure with exceptions. Existing symbolic and related connectionist research are described, and their limitations summarized. The requirements for an adaptable connectionist model are laid out, and a representational architecture is constructed. The architecture requires relations (i.e. links between objects) to be bi-directional or directional, where the former is meant to capture those relations for which it is useful to have the inverse relation (e.g. `isa', `part-of'). The general assumption is that inferential distances are best captured by relying on representational similarities in the semantic features of tokens and types. Both the encoding mechanism and the decoding mechanism (for checking the uniqueness of the distributed representations) are described in detail. The representational architecture is implemented in recursive autoassociative memory. The model is successful, and future adaptation or handling multiple inheritance with exceptions is briefly explored.

This paper describes a connectionist system for representing and reasoning with multiple inheritance structures with exceptions. The representational architecture has three characteristics. First, it merges relational with taxonomic representations. Secondly, it handles conflicts generated by exceptions and the use of multiple superclasses. Thirdly, it uses fully distributed representations. One novel feature is that, since the distributed representation of an entity is influenced by its position in the inheritance structure, representations of assertions are influenced by the context of the entities. An extension to the model which implements and makes use of confluent inference is described.

A commonly used methodology in the design of a database system is to use the ER Model for the conceptual design together with the RDM during the implementation phase. This methodology requires that a mapping between the two data models be performed.

This paper addresses the `lost semantics' problem caused by the impedance mismatch between the two models and reports on a prototype system which has been developed in order to bridge this gap. The aim of the system is to provide a tool for preserving the semantics of a conceptually rich model when automatically translating it into a working system.

The graph oriented approach suggested by Dogac [Dog90] makes use of the implicit information from the ER Schema when identifying the correct update propagations. They suggest a preprocessor which must be run on all applications for this automatic generation of relational code. We take this a stage further and suggest how, using the

emerging active technology of new DBMS, the techniques can be used to provide full, automatic and centralised support for an ER design. Details of an initial prototype (for non-active systems) and a refined prototype (for the latest release of a system, with triggers) are provided.

There are a number of aspects to the architecture which has been developed: it is multi-level, for targeting different database systems; it preserves the ER Model in the working system; it can be used for both active and passive database systems. The early (passive) prototype has been designed to run on top of the commercial RDBMS

INFORMIX. Work on both the passive and active designs targets the commercial RDBMS INFORMIX; the active design uses a general intermediate form of ECA rule as a target for translation.

HS-IDA-TR-93-003. Annotation: Published in The Proceedings of the 1993 Connectionist Models Summer School, (Eds) Mozer et al., Lawrence Erlbaum, 1993.

Connectionism as a model of the mind has been attacked by the advocators of the classical paradigm, who claim that Connectionism can only work if it is an implementation of Classical representations. This could be true for some of the models that claim to be Connectionist, but it will in this paper be shown that this is not true for Connectionist architectures that use non-symbolic representations. We will provide evidence in the form of simulation results that severely weaken of the arguments raised by Fodor and Pylyshyn and Fodor and McLaughlin, including their two main arguments, which are the lack of compositionality and systematicity.

Abstract: Connectionism as a model of the mind has recently been challenging the Classical model, in which the mind is regarded as symbol manipulating system. The main arguments against Connectionism concern its inability to form mental representations for complex expressions, which can be used for structure sensitive operations. Some argue for hybrid models which combine some of the most attractive features of the Classical and Connectionist models. This paper starts off by examining the definitions of the different approaches and also their strengths and weaknesses. One section is devoted to the debate between the advocators of the different paradigms, including the arguments about the lack of compositionality and systematicity in Connectionist cognitive models. We then argue for the Connectionist approach as the most attractive model of the mind. This includes performing the "miracle" of defining structure sensitive operations on non-symbolic representations of concepts.

It has been argued that models, that are claimed to be models of the mind, have to exhibit a behaviour closely related to human thought. This includes dealing with the issues of compositionality, systematicity and productivity. This paper starts by describing a non-concatenative mode of combination for connectionist patterns of neural activation. We then turn to the issue of systematicity, i.e. structure sensitive processes. We explore this issue to some level of detail, e.g. the importance of choosing the `right' type of representation and how the construction of the training set could result in different types of systematicity.