Information Technology Reference
In-Depth Information
Fig.9.
Number of correct classification as a function of inactivity probability. The
curves refer to the GKL rule and to two asynchronous CAs.
Once again, the results of this section confirm that asynchronous CAs degrade
much more gracefully than synchronous ones in noisy environments. They thus
intrinsically offer more resilience and robustness.
5 Conclusions
In this work we have shown that physically more realistic asynchronous CAs
of various kinds can be effectively evolved for the density task using genetic
algorithms, although their performance is lower than that obtained by evolved
synchronous CAs. We have also shown that the computational strategies discov-
ered by the GA in the asynchronous case are different from those of synchronous
CAs due to the presence of a stochastic component in the update. This very
reason makes them more resistant to changes in the environment and thus po-
tentially more interesting as computational devices in the presence of noise. In
the same vein, they seem to have better scalability to larger computational grids
than evolved synchronous CAs. Other important aspects that we are studying
but are not included here are further investigations into their fault-tolerance
properties.
Acknowledgment.
We acknowledge the Swiss National Fund for Scientific
Researchfor financial support under the grant 21-58893.99.
References
1. H. Bersini and V. Detour. Asynchrony induces stability in cellular automata based
models. In R. A. Brooks and P. Maes, editors,
Artificial Life IV
, pages 382-387,
Cambridge, Massachusetts, 1994. The MIT Press.
