home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!ucbvax!bloom-beacon!INTERNET!dont-send-mail-to-path-lines
- From: hurd@math.gatech.EDU (Lyman Hurd)
- Newsgroups: comp.theory.cell-automata
- Subject: What is Garden of Eden?
- Message-ID: <9211072208.AA09554@math.gatech.edu>
- Date: 7 Nov 92 22:08:35 GMT
- References: <1992Nov2.155048.2826@hellgate.utah.edu>
- Sender: root@athena.mit.edu (Wizard A. Root)
- Distribution: inet
- Organization: The Internet
- Lines: 48
-
-
- Date: 2 Nov 92 15:50:48 MST
- From: tolman%asylum.cs.utah.edu@cs.utah.edu (Kenneth Tolman)
- Organization: University of Utah, CompSci Dept
- Newsgroups: comp.theory.cell-automata
- Sender: ca-request@Think.COM
-
- What is the Garden of Eden? What was Moore's theorem about these
- configurations?
-
- I have been assuming that a Garden of Eden configuration is some state which
- can only exist during the first time interval of a CA.... such as a
- grid of entirely on cells for the game of life. Is this a correct
- intepretation? Are there important implications for these configurations?
-
-
- This is correct. Specifically if the state space of the CA is X and the
- rule represented by a global function F: X --> X, then a configuration c in
- X is a GOE if and only if c in X - F(X).
-
- From my own persppective, what are interesting are the complements of the
- GOE's. or a physical system, it seems reasonable to assume that it is in
- steady state and therefore one assumes that even if the dynamics are
- irreversible one can assume that an observed state has a semi-infinite
- sequence of predecessors (by no means unique).
-
- In one dimension, the GOE's for a given iteration (those configurations
- which can occur at time step N but not at time step N+1) collectively form
- a recursively enumerable set (in a sense made precise in the paper cited
- below but which can be explained if asked).
-
- In [1] showed that the set need not be recursive. Hence it was undecidable
- whether a given configuration was GOE for ALL n although it IS decidable in
- 1-D for a specific N. The question in 2-d is actually somewhat easier
- because less is true. The undecidability of the Tiling problem (Berger's
- Theorem) has wide-ranging implications for 2D CA.
-
- I thought I would throw this in since I assume that my perspective is
- fairly peculiar in this regard. The notation comes from Moore's Garden of
- Eden Theorem for which I hope someone will jump in with a reference. Also
- there are some papers of Golze which generalize the problem in several
- ways. Once again I will try to dig up references but other people shoould
- feel free to jump in.
-
- Hope this helps,
-
- Lyman Hurd
- Iterated Systems, Inc.
-