A New Adaptive Fast Cellular Automaton Neighborhood Detection and Rule Identification Algorithm.

Area of Science:

• Complex Systems
• Computational Science
• Artificial Intelligence

Background:

• Cellular Automata (CA) identification requires accurate neighborhood detection before parameter estimation.
• Existing methods using neighbor removal can lead to ill-conditioning and overfitting, especially with large initial neighborhoods.

Purpose of the Study:

• To propose a novel Adaptive Fast Cellular Automata Orthogonal-Least-Square (Adaptive-FCA-OLS) algorithm.
• To enhance neighborhood detection accuracy and reduce computational demands in CA identification.

Main Methods:

• Introduction of a new criterion and three novel techniques for adaptive neighborhood searching.
• Development of the Adaptive-FCA-OLS algorithm to avoid preset tolerance in neighborhood detection.

Main Results:

• The Adaptive-FCA-OLS algorithm adaptively searches for the correct neighborhood.
• Demonstrated reduction in computational complexity and memory usage.
• Improved robustness to noise and varying initial neighborhood sizes compared to existing methods.

Conclusions:

• The Adaptive-FCA-OLS algorithm offers a more robust and efficient approach to identifying binary CA.
• This method addresses limitations of traditional neighbor removal techniques in CA modeling.