Showing papers on "Artificial neural network published in 1971"

Adapting coincidence scalers and neural modelling studies of vision.

Abstract: Some extensions of the theory of adapting coincidence scaling are presented in the context of neural theory and modelling. Previously the theory of adapting coincidence scaling has been successfully applied to quite a number of specific problems mainly drawn from psychophysical theories of vision: van de Grind et al. (1970a, b); Koenderink et al. (1970a, b). Here emphasis is on neurophysiological problems and after a brief discussion of the “coding” and “component” problems of neural network modelling and a survey of basic coincidence scaling mechanisms a paradigm for neural encoding is treated in some detail. This paradigm (Fig. 6A) is similar to the neuromimes developed and studied by Harmon (1959, 1961) and Kupfmuller and Jenik (1961) for deterministic input signals. On the basis of the introductory discussion of the coding problem it is assumed that the neural code in the peripheral part of the nervous system that we choose as our hunting ground, viz. the retina, is an average event rate code with a Poisson point process as a carrier. Thus the paradigm for neural encoding is studied for such a stochastic input point process. It is then among other things shown that such a simple encoder can generate a wide variety of multimodal interval distributions for certain choices of its parameters. Next we turn to a classic coincidence model of vision and give extremely accurate simulation results to substitute for the lacking analytic solution of the underlying K-fold coincidence problem. A shortcoming of this model is analysed in terms of elementary neural operations and it is shown that the problem of specifying a generalized version of the model ties in with the problem of developing models to explain the quantal signals (bumps) observed on the generator potential during intracellular recordings from the eccentric cell of Limulus. A cybernetic principle for “bump” size adaptation is formulated on the basis of the apparent and possibly significant similarity of this adaptation process with the event rate reduction principle embodied in the so called V R-machine (van de Grind et al., 1970a) which is one of our set of adapting coincidence scalers.

Natural pattern clustering in digital learning nets

Abstract: This letter reports the results of experiments in which an unsupervised net is found to perform a natural-clustering operation. An `aging? process is introduced during learning, and the net is randomly connected with outputs feeding back to inputs.

Effective computation over the real numbers

Abstract: Extended Turing Machines We will obtain our simplest model, the ETM, by adding to Turing Machines the ability to add, multiply, and compare real numbers. We allow real parameters to appear in an ETM's program. In this manner the set of real numbers becomes the underlying domain of our computations much as a finite set of symbols is the domain of computation of a Turing machine. As successively deeper levels of GLB are allowed, successively stronger computational systems result. Functionals which can be approximated iteratively at one level (such as differentiation) can be computed exactly at a somewhat higher level. By closing our computation scheme under the GLB-operation we obtain a system closed under limiting operations in general, but one for which the question of effectively choosing elements from effectively defined sets is presently unanswered. Indeed, the solution to this last question may depend upon higher axioms of set theory. More explicitly, take as the definition of an (ordinary) Turing machine a device consisting of (1) an infinite tape divided into squares on which symbols from some finite alphabet are written. (2) a read-write head capable of moving one square left or right under the control of (3) a finite-state-control (fsc). We consider an fsc to be a flowchart consisting of an allowable interconnection of boxes of the following forms: multiply number on tape with that in s and store result in s branch according to sign of number in s the real number beneath the head is copied into s the real number in s is written on square of top track currently scanned by the head (for c some real number) the constant c is put into s add number on top track square currently scanned to that in s and store result in s (j) s+T-+ s (i) c-+ s (1) Test s (k) s*T-+ s (h) Store s (g) Load s flowchart Any number of arrows may enter a HALT box, but none leave it (c) TEST This box has 1 outgoing arrow for each symbol in the Turing machine's alphabet. TEST causes a branch according to the symbol being scanned; if the ith symbol of the alphabet appears beneath the head when ~ TEST box is entered , control passGs to the flow-chart box indicated by the ith arrow leaving the TEST box (d) Write symbol i The ith symbol of the Turing machine's alphabet …

Unit spike activity in the coelenteran neural network

Abstract: Umgekehr t verkfirzen tiefere Umgebungs tempera turen die n&chtliche Aktivit&t erheblich. Unter t 5 ~ Umgebungs temperatur setzt Dauerlethargie ein. Die Ergebnisse lassen bereits klar die Stellung der Tanreks unter den hetero thermen Insektivoren erkennen. Sie sind stets zn ether Senkung der K6rper temperatur bereft, wenn dies die Umgebungs tempera tu r erfordert. Ihre Fettspeicher erm6glichen auch tagelange Dauerlethargie ohne Nahrungsaufnahme. Eine genaue Analyse der Temperaturregulat ion unter verschiedenen Umgebungsbedingungen wird gegenw~rtig durchgeffihrt.

A note on a model for the quasilinear wave equation

Abstract: at those points (x, t) where Z is differentiab[e. We shall answer these questions for an equation which models (E). To motivate the choice of the model equation we are forced to say a few words about the nature of the unknown Z. The function Z(x, t) represents the position at time t of a material point which was located at x at t=0 . Similarily, Z~(t) gives the position at time t of a particle (of mass h) which was located at kh at t=0 . We envision these particles as moving on a line and therefore insist that the following inequalities be satisfied for all

Learning process in a model of associative memory

Abstract: The excellent information processing in a human brain is considered to depend upon its association mechanisms To simulate this function, we propose in this paper a model of the neural network named “Associatron” which operates like a human brain in some points Associatron stores many entities at the same place of its structure, and recalls the whole of any entity from a part of it From that mechanism some properties are derived, which are expected to be utilized for human-like information processing After the properties of the model have been analyzed, an Associatron with 180 neurons is simulated by a computer and is applied to simple examples of concept formation and game playing Hardware realization of an Associatron with 25 neurons and thinking process by the sequence of associations are mentioned, too

On the design of easily testable sequential machines

Abstract: In order to obtain a short fault-detection sequence for a sequential machine, the concept of an easily testable machine is introduced. Such a machine is one which possesses a minimal-length homogeneous distinguishing sequence and requires no transfer sequences in the fault-detection sequence. A design procedure is presented in which an arbitrary machine is embedded in an easily testable machine by adding input lines to the original machine. The procedure also derives a fault-detection sequence for the easily testable machine.