Autosophy Networks yield self-learning Robot Vision

Replacing programmed vision systems with self-learning vision systems

Authors: Klaus Holtz

Published / Presented at: Wescon/93, Moscone Convention Center, San Francisco, California, September 28 – 30, 1993.

Level: Expert: Vision Systems

Abstract: Autosophy is a new science, which tries to explain self-assembling structures, such as crystals or living trees, in mathematical terms. This research provides algorithms for growing self-learning data networks in memories, similar to the growing of crystals or data trees, without programming or conventional data processing. The new learning algorithms provide an entirely new technology for building brain-like robot vision systems. Television images from a camera are learned and stored in a peculiar omni dimensional hyperspace library. This will result in enormous data compression because no portion of the image is stored twice. The more images have already been learned the fewer storage locations will be required to store additional images, The omni dimensional hyperspace library allows for near instant search access, which is independent of the library size. The resulting brain-like robot vision system are educated very much like a human child by showing objects to a television camera and associating them with descriptive text or robot arm motions. Autosophy may succeed where conventional data processing has failed in providing intelligent self-organizing robot vision. Text and image learning has been simulated in software to show learning behavior, which is strikingly similar to our own brain.

Available downloadable documents:

Publication Top Page -- Adobe pdf

Published document -- Adobe pdf