Israel Lectures ,
Sept. 27 to Oct.6, 2005
Hebrew University of Jerusalem (HUJI), Givat Ram Campus;
Weizmann Institute of Science, Rehovot;
Bar-Ilan University Tel Aviv, Ramat Gan Campus.
25/08/2005
Introduction to Interference Networks
Gerd Heinz, GFaI Berlin, info@gheinz.de
Overview
Applying physical parameters, like velocity, spike-like time functions and spherical arrangements to neural (weight- or pattern-) networks, they change their properties in a fundamental way. We will call them Interference Networks (IN).
Input locations appear mirrored on the output field - a known phenomenon from bio-medical researches. Adjacent neurons cannot excite one another, nets have to be cleaned before they work, slowness increases memorized capacity and location sensitivity and learning occurs on defined points only. Even elementary functions of neurons change. In this new world everything changes. The behavior of such nets covers properties of known nerve experiments, like Singers synchronization, Penfields "Homunculus", the zoom lens of visual attention, pain, auditory maps or somato-topic maps.
New questions appear:
By physical means IN appear as genus over code- and frequency- selective and projective networks. Possible applications range from optical lens systems, acoustic imaging, digital filters, frequency maps, antenna pattern up to classic neural nets (as a special class of time-restricted networks).
Investigations of IN can help to define possible directions of experimental research.
As an illustration how somatotopic cards work, years before I created the Acoustic Camera technology, which got different awards within Germany, see http://www.gheinz.de
Here the image pixels work comparable to interfering neurons.
Contents
0 Introduction
Looking behind Jeffress model of sound localization
Example for mirror property
Thumb experiment
Andrew’s squid experiments
Acoustic Camera example
1 Impulse-Waves on Wires
Consequences of the integrate-and-fire-principle
Waves on short- circuited nets of wires
The wave- field-of-function concept
Multiplication und addition of time functions
Correlation of monopolar signals
Equation extraction
Reflection and refraction
2 Interference Circuits (IC)
Simplest ICs
Neuronal basic functions
Projective ICs
Superposition: normal and "eating" waves
Inhibition by excitement
Examples
3 Cross- and Self-Interference Properties
Projections
Code maps
Frequency maps
Filter-theoretical analogue
4 Projective Properties
Zooming und movement (scaling)
Conjunction and permutation
Permutations in somatotopic and autonomous fields
N-dimensionality (over-conditioning)
Reconstructing pseudo- wave fields
5 Projection Circuits
Colored interference systems
Space-time-interlock
Declined maps
h-, e-, p-, k- inversions
6 Calculating simple ICs
Mask algorithm
Reconstruction contra projection
Channel number and cross- interference distance
Cross- interference overflow (pain?)
Properties for somato-topic resolution
Inter-medial projections (dolphin, bat)
7 Discussion of IC-Examples
Noise maps
Mirrors
Scaling
Memorized volume
Homunculus
Visual Cortex
Equilibrium circuits, rule of glia
Singer’s synchronization
8 Demonstration
Acoustic image
Acoustic movie
Frequency images and movies
Source separation
___________________________
Speaker
Dr. G. Heinz, GFaI (Society for the Promotion of Applied Informatics)
Rudower Chaussee 30, 12489 Berlin
Tel.: +49 30 6392-1652/ -1600, Fax: -1602
info@gheinz.de
http://www.gheinz.de/
http://www.acoustic-camera.com
"After 3 books, 80 patents and 130 publications I do no longer believe in the contents of the written word." Michael Goessel, 1990