home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!paladin.american.edu!howland.reston.ans.net!zaphod.mps.ohio-state.edu!rpi!ghost.dsi.unimi.it!insa-lyon.fr!univ-lyon1.fr!frmop11!barilvm!technion!haifauvm!rsma410
- Organization: University of Haifa - Mt. Carmel Haifa, Israel.
- Date: Monday, 25 Jan 1993 15:45:49 IST
- From: Dan Gordon <RSMA410@HAIFAUVM.BITNET>
- Message-ID: <93025.154549RSMA410@HAIFAUVM.BITNET>
- Newsgroups: comp.theory.cell-automata
- Subject: Re: cellular automata with different rules at each site
- References: <C188rz.2Jq0@austin.ibm.com>
- Lines: 49
-
- In article <105853@netnews.upenn.edu>, bill@amy.med.upenn.edu (Bill Tozier)
- says:
- > Is anyone familiar with any work that's been done with
- >cellular automata in which the rules for any given cell are fixed,
- >but may be different for different cells? I'm currently looking
- >(on the back of an envelope, literally) at behavior of 1D, neighborhood-3
- >2-state automata when each cell has a random rule table. Without
- >much stretching, this type of automaton would seem to be a re-description
- >of a class of networks that includes Boolean nets (a research
- >specialty in these parts).
-
- It's not clear what you mean by "random rule table". This can be
- interpreted as a probabilistic automaton. It can also be interpreted
- as meaning that in every cell, the rule table is fixed (and not
- probabilistic), but the choice of which rule table goes in which
- cell is random. A consequence of this is that more than one rule
- table can appear infinitely many times in the initial configuration.
-
- Assuming you are referring to the second possiblilty, the answer is
- simple: There are only 256 such rule tables (or transition functions)
- so your automata can be simulated by homogeneous automata with 512
- states, in which the initial configuration is chosen randomly and
- infinitely many cells may be in a non-quiescent state.
-
- Further explanations: a rule table (in the type of automaton that
- you use) is a function from the set of all triples of 0,1 to the
- set {0,1}, and there are 256 such functions - call them T_1...T_256.
- The homogeneous automata that will simulate these will have states
- of the type (A, i), where A is in {0,1} and 1 <= i <= 256; a total
- of 512 states. The rules for the homogeneous automata are as follows:
-
- If a cell is in state (A, i), its left neighbor in state (B, j),
- and its right neighbor is in state (C, k), then simply apply rule
- T_i to the 3-neighborhood B-A-C. Suppose rule T_i changes state A
- to state D, then the simulating automaton should change the current
- state (A, i) to state (D, i). Note that i does not change.
-
- The simulation is now straightforward: a cell in state A, having rule
- table T_i is simply simulated by a homogeneous automaton in state (A, i).
-
- The usual assumption in cellular automata is that in the initial
- configuration, all but a finite number of cells are in a special state
- called the "quiescent state". However, in your automata, you allow
- more than one rule table to appear infinitely many times, so the
- simulating automata must also be allowed to have more than one state
- appear infinitely many times. I believe this has been looked at
- in the literature, but I have no references available.
-
- Dan Gordon.
-