Discrete Dynamics Lab

Nov 2012

ddlabx11 New options for NULL boundary conditions

Supersedes ddlabx09, June 2012 (also ddlabx10) and includes further changes since the publication of Exploring Discrete Dynamics in May 2011

NULL Basin of attraction field, ECA rule 150, n=11
DDLab has been updated at regular intervals since its release in 1995. Its precursor was the Atlas software included on diskette inside the back cover of "The Global Dynamics of Cellular Automata".
For a list and download of this and older versions click here.
Below are links to descriptions of previous updates,
June 2012
Nov 2005
Dec 2003
July 2001
Feb 1999
Sept 1997

download ddlabx11 for Linux, Mac, Cygwin and DOS
summary of updates (EDD#x.x.x refers to the relevant chapter or section in "Exploring Discrete Dynamics")

examples -- click to enlarge

ECA rule 225 n=14 - Periodic boundaries

ECA rule 225 n=14 - Null boundaries

ECA rule 110 n=4 to 11 - Null boundaries

ECA rule 110 space-time patterns n=150
Null boundaries. Cells colored by neighborhood lookup EDD#32.9.5

ECA rule 52 n=4 - Null boundaries [3]

mixed ECA rules 105,177,171,75 - Null boundaries [4]

mixed ECA rules 5,73,200,80 - Null boundaries [5]

mixed ECA rules 10,69,204,68 - Null boundaries [5]

Null Boundary Condition for Cellular Automata and Discrete Dynamical Networks

DDLab has been extended for Null Boundary Conditions (NBC), where inputs beyond the network's edges are held at a constant value of zero (though another value is possible). This contrasts with Periodic Boundary Conditions (PBC), up till now the only method adopted for cellular automata (CA) in DDLab.

NBC are of interest in pattern recognition, and other applications where the system is grounded or quenched, or bounded by edge, skin or membrane. As for PBC, NBC are also interesting as mathematical/dynamical systems in their own right.

All DDLab functions and options for computation, display and analysis, as described in the book "Exploring Discrete Dynamics" (EDD), can now also be applied to NBC systems, multi-value as well as binary. This includes CA in various dimensions, but also random Boolian networks (RBN) and discrete dynamical networks (DDN), because irrespective of the wiring scheme, for NBC any part of the pseudo-neighborhood that extends beyond the network's edges is made to take zero as its input.

NBC space-time patterns (running forward) apply to 1d, 2d and 3d networks -- toggling between NBC and PBC on-the-fly. All (forward) functions and methods apply -- such as the look-up entropy, filtering, damage, and attractor histograms.

NBC basins of attraction (running backward) apply to 1d networks by means three completely different reverse algorithms, where the original PBC algorithms have been modified for NBC: (1) for CA, (2) for RBN/DDN as well as CA, and (3) the exhaustive reverse algorithm for any of the above, thus providing a reality check of results. All (backward) methods and functions for basin of attraction fields, single basins and subtrees apply.

The new NBC functionality in DDLab opens up a new area of phenomenology (as before for PBC) for quick and easy investigation. Early results show that NBC basins of attraction are very different from PBC. There is a greater chance of point attractors or short periods. For CA with symmetric (odd size) neighborhoods (k=3,5,7,...) the reflection transformation gives equivalent dynamics/basins. Space-time patterns (including gliders) exhibit the same dynamics away from edges, of course, but mobile structures are usually quenched at the edges (resulting in shorter cycles and taransients), or may sometimes interact (bounce) with the edges.

How to set Null Boundary Conditions in ddlabx11

running forward:

At the prompt in EDD#31.1 there is an extra option: boundaries-b. If 'b' is entered, there is a further prompt (initially): PERIODIC boundaries, set NULL-N: or NULL boundaries, set PERIODIC-P: A change (to NULL) will be conserved unless backtracking to the seed prompt, EDD#21.1.
While space-time patters are running, PERIODIC/NULL can be toggled on-the-fly with key hit '|' (vertical bar symbol). The current status appears in the "on-the-fly key index" EDD#32.1, either "Periodic/NULL" or "NULL/Periodic"

running backward:

At the prompt in EDD#24.1 there is an extra option attached to the existing "display" category, so now it reads ... display/boundary-p
If 'p' is entered, the next prompt EDD#26.1 shows the current setting, PERIODIC initially, and an option to change to NULL ...
PERIODIC boundaries (compression ON, off-s:) NULL-N:
or vice-vera ...
NULL boundaries (compression OFF) PERIODIC-P:
A change (to NULL) will be conserved unless backtracking to the seed prompt, EDD#21.1 (for single basins and subtrees) or (for a basin of attraction field) to the neighbourhood size prompt EDD#9.1.
PERIODIC or NULL appears in the top-left corner of the "attractor basin complete data window" EDD#27.2.

Attractor basins can be generated using three alternative reverse algorithms selected in EDD#29.4. An acronym for the algorithm used now appears in the top-right corner of the "attractor basin complete data window" EDD#27.2, as follows ...

CAW ...CA wiring algorithm, the 1d direct algorithm, which allows mixed rules.
NLW ...non-local wiring algorithm, for CA, RBN, DDN, which allows mixed k.
EXH ...exhaustive algorithm #29.7, for CA, RBN, DDN (required for MAP, and sequential updating).

The following acronyms are also added, but do not relate to NULL boundaries.
MAP ...random map #29.8, using the EXH algorithm (PERIODIC/NULL is irrelevant).
SEQ ...sequential updating #29.9, using the EXH algorithm (disabled for NULL).
NTO ...neutral order components #29.10, related just to wiring (disabled for NULL).

Confirming previous NBC results for CA in the literature:

Rule 150, n=7 and n=11, fig. 1 [1].
Rule 90, n=1 to 40, Table 1. Transient and Cycle Lengths [2] (checked by attractor basins and attractor histogram).
Rule 52, n=5, fig.1 [3].
Mixed rule CA 105,177,171,75, n=4, Fig.3.5 [4].
Mixed rule CA 5,73,200,80, n=4, Fig.1 [5].
Mixed rule CA 10,69,204,68, n=4, Fig.5 [5].

NBC references

1. N.Pitsianis, Ph.Tsalides, G.L.Bleris, A.Thanailiakis, H.C.Card "Deterministic One-Dimensional Cellular Automata" Journal of Statistical Phisics, Vol.36. Nos.1/2, 1989.

2. John G. Stevens, ~ Ronald E. Rosensweig, 2 and A. E. Cerkanowicz, "Transient and Cyclic Behavior of Cellular Automata with Null Boundary Conditions" Journal of Statistical Physics, Vol 73, No.1/1, 1993.

3. Sudhakar Sahoo, Pabitra Pal Choudhury "Issues on drawing the State Transition Diagram for arbitrary Cellular Automata", Applied Statistics Unit, Indian Statistical Institute, Kolkata, 700108, INDIA, (where published? 2008).

4. Sukanta Das, DPhil, 2006 "Theory and Applications of Nonlinear Cellular Automata In VLSI Design", Department of Computer Science and Technology, Bengal Engineering And Science University

5. Nazma Naskari, Sukanta Das, Biplab Sikdar "Characterization of nonlinear cellular automata having only single length cycle attractors" JCA, 2012 (to appear).

Return to the Discrete Dynamics Lab home page.
Last modified: Dec 2012