STRATEGIES OF FUNCTION DECOMPOSITION FOR ARTIFICIAL INTELLIGENCE, VOLUME II.
Abstract
Preliminary results are reported in 13 research notes on strategies of function decomposition solely from observations of inputs (variable configurations) and outputs (function values). The classes of functions to which the results apply include discrete, finite, deterministic functions as well as arbitrary close discrete approximations to continuous functions of continuous variables. Nondeterministic (i.e., probabilistic) and sequential (i.e., finite automata) functions are not included. The research notes consider (a) decomposition costs and the equivalence of all measures of cost or complexity; the detection of economical decompositions; and (c) generalizing properties of economical decompositions. Efficient procedures are suggested for detecting economical non-composite decompositions of any given partial or total discrete function solely from input and output observations. Composite decompositions become tractable when enough is known or properly conjectured about their sub-functions. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1965
- Accession Number
- AD0620186
Entities
People
- David G. Willis