By Topic

Symbols Versus Neurons, IEE Colloquium on

Date 1 Oct 1990

Filter Results

Displaying Results 1 - 11 of 11
  • Symbolic constraint-based reasoning in Pandora

    Page(s): 9/1 - 9/5
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (244 KB)  

    Introduces an intelligent programming technique to adopt constraint-based reasoning in Pandora: a non-deterministic parallel logic programming language. The technique is illustrated in solving resource allocation problems, such as automatically generating naval flying programmes View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • IEE Colloquium on `Symbols Versus Neurons' (Digest No.123)

    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (20 KB)  

    The following topics were dealt with: knowledge and machine architecture; systolic algorithms for back-propagation; connectionism; genetic algorithms; neural networks and MIMD-multiprocessors; real-time reinforcement learning control; relational and differential logic for knowledge processing; constraint-based reasoning in Pandora; subsymbolic inductive learning framework; genomic interpretation; and inductive protein structure analysis View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Artificial intelligence for genomic interpretation

    Page(s): 11/1 - 11/4
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (200 KB)  

    Biological entities present a complexity level that should be correctly managed in Artificial Intelligence environments. One has to describe and access them in a proper way. What is the role of symbolic learning in this context? The authors define semi-empirical knowledge and theories, which constitute their goals. Knowledge is represented by the means of conceptual graphs, which allow one to manage easily the control of any process of iterative learning to refine the expert's knowledge. They present an illustration of these principles in a specific domain, protein folding View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Real time reinforcement learning control of dynamic systems applied to an inverted pendulum

    Page(s): 7/1 - 7/2
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (108 KB)  

    Describes work started in order to investigate the use of neural networks for application in adaptive or learning control systems. Neural networks have learning capabilities and they can be used to realize non-linear mappings. These are attractive features which could make them useful building blocks for non-linear adaptive or learning controllers View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Knowledge and machine architectures

    Page(s): 1/1 - 1/9
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (412 KB)  

    A paradigm is proposed based upon a taxonomy of knowledge; a taxonomy that has been strongly influenced by the need to represent knowledge for machine processing. The importance of such a paradigm is to show an equivalence of activity in all spheres of system design from knowledge systems to machine architectures and thus open up the possibility of cross fertilization of techniques and a redistribution of tasks across fields. Both these possibilities will improve the design of knowledge systems within the framework of a rapidly evolving technology View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Subsymbolic inductive learning framework for large-scale data processing

    Page(s): 10/1 - 10/8
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (300 KB)  

    Recent years have witnessed the development of a large variety of Inductive methods for data analysis. This can be attributed to the fact that the decision tree-the most common representation of Inductive algorithms-provides a hierarchical framework for sequential decision making. This is a framework which non-professionals find easy to use and understand. Furthermore, it has been proved that Inductive Learning performs as well as, and indeed often better than Discriminant analysis and Multi Logic/Probit analysis. It has been also pointed out that some problems such as protein structure prediction, which are unsolvable with statistical methods can be approached quite successfully with Inductive methods. The authors aim in the paper is to express their experience in Inductive Learning in a strict form. They call this approach the subsymbolic Inductive Learning Framework, because it explores very primitive syntactic objects, and builds from them compound knowledge structures View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural networks and MIMD-multiprocessors

    Page(s): 511 - 514
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (536 KB)  

    Two artificial neural network models are compared. They are the Hopfield neural network model and the Sparse Distributed Memory model. Distributed algorithms for both of them are designed and implemented. The run time characteristics of the algorithms are analyzed theoretically and tested in practise. The storage capacities of the networks are compared. Implementations are done using a distributed multiprocessor system View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Inductive protein structure analysis (IPSA)

    Page(s): 12/1 - 12/7
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (400 KB)  

    The current state of protein structure analysis and prediction methods shows three important points. First, classical methods of secondary structure prediction cannot solve the protein folding problem; secondly, a combination of classical method returns better results for the prediction of secondary structures than any one of the methods on its own; and thirdly, methods that try to incorporate the effects of long-range interactions produce a better set of results than comparable methods not using this information. This is the starting point and motivation for a new method of computer-assisted protein structure analysis. The name `Inductive Protein Structure Analysis (IPSA)' indicates the crux of the method, which is the automated search for patterns and structural regularities at different levels of the structure of proteins. The concept of `induction' generally describes the process by which a rule is derived from a set of examples View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Using the genetic algorithm to adapt intelligent systems

    Page(s): 4/1 - 4/4
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (180 KB)  

    The genetic algorithm, loosely based on the mechanics of evolution, is used in machine learning and optimisation problems that typically have a large search space and require a high tolerance to noise. Two examples are given of its use in the learning of rules for real-time control problems; one for adaptive rule-based optimisation of combustion in multiple-burner installations in the steel industry and the other for controlling a dynamical system. Current research on genetic algorithms is largely focussing on their use for optimising neural networks, since this is a natural way of combining the paradigms of evolution and learning, and on parallel and distributed implementations, to facilitate the efficient solution of larger problems. A project using a parallel implementation of an incremental genetic algorithm to generate constraint networks from raw data is described View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Connectionism-a link between psychology and neuroscience?

    Page(s): 3/1 - 3/7
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (512 KB)  

    `Neural network' or connectionist models are currently `in' in psychology and cognitive science. Why is this so? The author of this paper thinks that one important reason for this is the hope of many psychologists and cognitive scientists that by using such models the gap between theories of the mind and behavior on the one hand and theories of the brain on the other hand could be made narrower. If this was really the case, then adopting connectionism would be a major step forward for psychology. But is this hope substantiated by the results of the research on neural networks? The aim of this paper is to give a partial answer to this question View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A systolic algorithm for back-propagation: mapping onto a transputer network

    Page(s): 2/1 - 2/4
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (144 KB)  

    The paper is devoted to the implementation of Back-Propagation (BP) on local memory multiprocessor systems (LMM). First, a systolic algorithm (SA) is described, where dependencies are considered at the data item level. Next, the systolic array is partitioned and mapped onto a multiprocessor system. At this stage, the level of granularity is increased, in order to reduce communication cost. Finally, each stage is implemented on a transputer based multiprocessor, and their performance is compared with a simple sequential version of the learning rule. A parallelization rate of about 0.9 is obtained. Back-Propagation is a supervised, gradient descent learning rule for multilayered, feed-forward connectionist networks View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.