Friday, May 4, 2012
Asa H 2.0's ancestral descent
Asa (autonomous software agent) is my AI project which began in about 1994-95. Asa F 1.0 was built from artificial neural networks (in an architecture like that of Richard Sutton's thesis, 1984) and early results were published in Trans. Kansas Acad. of Sci., vol. 100, num. 3-4, pg 85, 1997. Results from a version built from case-based reasoners, Asa F 2.0, were presented in Trans. Kansas Acad. of Sci., vol. 102, num. 3-4, pg 117, 1999 and Trans. Kansas Acad. of Sci., vol. 107, num. 1-2, pg. 32, 2004. A hierarchical version, Asa H 1.0, was described in Trans. Kansas Acad. of Sci., vol. 109, num. 3-4, pg. 159, 2006. The current Asa H 2.0 versions (see my 2011 book "Twelve Papers") employ revised representations (see my 22 Nov. 2010 blog) and have made use of parallel compution. An open source "lite" version of Asa H 2.0, with a fair amount of documentation, is presented in my 10 Feb. 2011 blog. Many experiments with these Asa packages have been reported in my various publications since 1995.