|
|
|
|
|
|
|
Keywords
Fundamental neurophysiological principles, advanced key issues, electronic realizations of
time-continuous and discrete neuron models
Details
Understanding Diagrams
Diagrammatic notations, such as state charts or musical notation, are an important component of human communication. Two of the most fundamental questions one can ask about such visual languages is how they may be specified and how to recognize and understand diagrams in a particular visual language. In this presentation we survey the many formalisms which have been suggested over the last three decades for visual language specification, discuss computational approaches to diagram understanding based on these formalisms and indicate possible applications. We shall also review recent directions in visual language theory, notably efforts to develop an analogue of the Chomsky hierarchy for visual languages, the specification of diagrammatic reasoning, and cognitive models of
visual language understanding.
From Control to Play, from Play to Construction
In this talk I will introduce the audience to the new results of R&D in Linguistic Geometry (LG). LG is a mathematical model for knowledge representation and reasoning about large-scale multiagent systems. A number of such systems including air/space combat, robotic manufacturing, software re-engineering, Internet cyberwar, etc. can be modeled as abstract board games. The purpose of LG is to provide strategies to guide the participants of a game to reach their global goals. Traditionally, finding such strategies required searches in giant search trees. Such searches are often beyond capabilities of modern and even conceivable future computers. LG dramatically reduces the size of the search trees, thus making the problems computationally tractable. Essentially, LG replaces search by construction of strategies that emerge implicitly when the agents (bombers, space interceptors, etc.) pursue their local goals by participating in the local skirmishes. A number of prototypes of LG systems and commercial products were developed at Lockheed Martin Corp., GIS Solutions, Sandia National Laboratories, US Air Force Phillips Laboratory, Rockwell International, University of Denver, and University of Colorado at Denver. Currently, STILMAN Advanced Strategies, Rockwell Science Center, and two universities, University of Colorado at Denver and Wayne State University, are working on the $1.5 million project on application of LG to "Agile Symbolic Mission Control and Hostile Counteraction" for the Defense Advanced Research Projects Agency (DARPA).
I will outline my new book "Linguistic Geometry: From Search to Construction" published by Kluwer Academic Publishers in March of 2000 and demonstrate JEC, a software prototype of an intelligent adviser to the Joint Chiefs of Staff with 2D/3D animated air force combat simulator.
Abstraction Levels in Neurocomputation-Properties of Physical and Mathematical Approaches
While the mathematical neurocomputation uses states, weights and patterns, the introduction of nerve fibre velocities and delays produces very different properties of the same network. Basing on fundamental research to wave interference in inhomogeneous, wiring interference systems (nerve-like) we will discuss the relevance of physical wave properties for informational purposes in nervous system.
Introducing (spheric, stochastic) delays and spiking time functions we will discuss basic 'neural' pattern networks (perceptron, nets for self organization, for association, for concept formation) for their ability to model bio-relevant behaviour under the aspect of real, slow velocities of fibres, of non-zero delays and of a non-zero volumes of the cells. We will find, that some pattern-carrying nets do not work correct under bio-physical conditions.
Using a wave theoretical approach we will discuss new basic functions of neural circuits in space and time. Simple physical models generate new properties: mirroring projections, moving, zooming, overflow. With the help of some simulation experiments with time-functions in space we will give some impressions about possible properties of neural assemblies. We introduce an informational role for nerve delays, wiring shapes, velocities, waveforms and channel numbers into a new field 'Wave Interference Informatics' and show, that the role of fibre delays is mutually more important as the role of synaptic weights. Concluding we will find out, that nerve behaviour is characterizable only by 'interference' networks, while 'neural' networks got more a technical disciplin.
A look in the history of neurocomputation shows the reasons for this development.