Showing papers on "Artificial neural network published in 1983"

Neuronlike adaptive elements that can solve difficult learning control problems

Abstract: It is shown how a system consisting of two neuronlike adaptive elements can solve a difficult learning control problem. The task is to balance a pole that is hinged to a movable cart by applying forces to the cart's base. It is argued that the learning problems faced by adaptive elements that are components of adaptive networks are at least as difficult as this version of the pole-balancing problem. The learning system consists of a single associative search element (ASE) and a single adaptive critic element (ACE). In the course of learning to balance the pole, the ASE constructs associations between input and output by searching under the influence of reinforcement feedback, and the ACE constructs a more informative evaluation function than reinforcement feedback alone can provide. The differences between this approach and other attempts to solve problems using neurolike elements are discussed, as is the relation of this work to classical and instrumental conditioning in animal learning studies and its possible implications for research in the neurosciences.

Neocognitron: A neural network model for a mechanism of visual pattern recognition

Abstract: A recognition with a large-scale network is simulated on a PDP-11/34 minicomputer and is shown to have a great capability for visual pattern recognition. The model consists of nine layers of cells. The authors demonstrate that the model can be trained to recognize handwritten Arabic numerals even with considerable deformations in shape. A learning-with-a-teacher process is used for the reinforcement of the modifiable synapses in the new large-scale model, instead of the learning-without-a-teacher process applied to a previous model. The authors focus on the mechanism for pattern recognition rather than that for self-organization.

Analog circuitry for the van der Pol and FitzHugh-Nagumo equations

Abstract: An analog circuit for the FitzHugh-Nagumo equations is given that uses readily available integrated circuitry. The mathematical model for this circuit is derived, and simple analysis is given, to show how the circuit works. Specifications for a reliable and easily built analog neuron are given with components that cost only a few dollars. The circuit is useful as a research and educational device.

Machine Learning: A Historical and Methodological Analysis

Abstract: Machine learning has always been an integral part of artificial intelligence, and its methodology has evolved in concert with the major concerns of the field. In response to the difficulties of encoding ever-increasing volumes of knowledge in modern AI systems, many researchers have recently turned their attention to machine learning as a means to overcome the knowledge acquisition bottleneck. This article presents a taxonomic analysis of machine learning organized primarily by learning strategies and secondarily by knowledge representation and application areas. A historical survey outlining the development of various approaches to machine learning is presented from early neural networks to present knowledge-intensive techniques.

Computing the Topological Entropy of Maps

Abstract: We give an algorithm for determining the topological entropy of a unimodal map of the interval given its kneading sequence. We also show that this algorithm converges exponentially in the number of letters of the kneading sequence.

Bifurcations in discrete dynamical systems with cubic maps

Abstract: A representative class of discrete dynamical systems with maps with two critical points is investigated numerically. In the symmetrical and the asymmetrical case a new type of bifurcation which does not occur for maps with one critical point is described. The corresponding reverse bifurcation is also found. New formulas for reverse bifurcation points, boundary curves and accumulation curves are given. 1. Introduction. Nonlinear discrete dynamical systems occur in many contexts in biological, economical and social sciences. A review with numerous references is presented in (10). The simplest system which is described by a first-order difference equation with a nonlinear map with one critical point has been investigated thoroughly by numerous authors (14), (13), (5), (3), (8) and (2). Universal metric properties for the bifurcations in this system have been discovered (4). The next most complicated map with two critical points has been considered in the symmetrical case by May (11) and (12) who pointed out that this system occurs in genetic problems where the selective forces depend on the gene frequencies. For the ignition phenomenon in neural networks a simple description giving rise to maintained activity has been presented in (15) and (1). The introduction of inhibitory connections in the network combined with refractory mechanisms may lead to systems with mapping functions with two critical points. In the present work we study bifurcations for systems described by first-order difference equations with a nonlinear two-parameter family of maps with two critical points. We consider symmetrical as well as asymmetrical cases. The bifurcations for systems with simple behavior characterized by solutions of low periodicity we denote forward bifurcations. In contrast reverse bifurcations (9) denote bifurcations in systems with complex behavior characterized by solutions of high periodicity or sensitive dependence on initial conditions. The latter type has recently been thoroughly investi- gated in a monograph by Collet and Eckmann (2). In ? 2 we formulate the model containing two parameters and recapitulate useful concepts for difference equations. In ? 3 we summarize the four types of forward bifurcations occurring in the model of which one appears not to have been treated in the context of difference equations. For certain types of reverse bifurcations occurring in the model we formulate new equations which for example determine the parameter values for the bifurcations. Section 4 contains our numerical results in terms of bifurcation diagrams obtained by solution of the governing equations as well as by numerous iterations. Regions in the parameter plane with different behavior of the system are mapped out. 2. First-order difference equation with a cubic map. First we shall summarize some basic facts concerning first-order difference equations with nonlinear maps. We

Machine Learning. Part 1. A Historical and Methodological Analysis.

Abstract: : Machine learning has always been an integral part of artificial intelligence, and its methodology has evolved in concert with the major concerns of the field. In response to the difficulties of encoding ever-increasing volumes of knowledge in modern Al systems, many researchers have recently turned their attention to machine learning as a means to overcome the knowledge acquisition bottleneck. Part 1 of this paper presents a taxonomic analysis of machine learning organized primarily by learning strategies and secondarily by knowledge representation and application areas. A historical survey outlining the development of various approaches to machine learning is presented from early neural networks to present knowledge intensive techniques. Part II (to be published in a subsequent issue) will outline major present research directions, and suggest viable areas for future investigation.