THEORY OF PROBABILITY STATE VARIABLE SYSTEMS.
Abstract
The simulation of the complete Neurotron on a digital computer is described in terms of flow diagrams, and flow diagrams for certain types of network iteration and goal action are presented. Simulated results show that (1) a single Neurotron can learn gain, and (2) show extensive statistical samplings of the behavior of a 36 Neurotron network when input signal changes are simulated. The mathematics upon which the analog functions of the Neurotron are based necessarily involve an approximation due to the digital nature of the computer, but it is shown that for analog time constants spanning more than ten computer cycles, the error is small. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1963
- Accession Number
- AD0428015
Entities
People
- R. F. Snyder
- R. J. Lee