A cellular automaton (CA) is a discrete computational system which consists of interconnection of cells that update their state at every time stamp according to some local rule. Rules are defined over neighborhoods and the most commonly used neighborhoods being Moore and von-Neumann neighborhood. In the past, cellular automata have been used in data classification tasks but the neighborhood definitions used are Moore and von-Neumann neighborhood, which are purely local. This paper proposes an efficient data classification algorithm known as Gaussian cellular automata (GCA) that uses cellular automata with semi local neighborhoods i.e., Gaussian kernel which includes Moore neighbors in the specified radius. Here, all neighbors do not have equal contribution in determining the label, their contribution decreases with increase in distance and vice versa. Experiments show that the proposed algorithm performs better than the existing cellular automaton models, i.e., Fawcetts Model [2] and Omers models [3] in terms of accuracy, execution time and number of generations required for convergence.
[1]
Palash Sarkar,et al.
A brief history of cellular automata
,
2000,
CSUR.
[2]
Peter Kokol,et al.
Building Classifier Cellular Automata
,
2004,
ACRI.
[3]
David W. Aha,et al.
Instance-Based Learning Algorithms
,
1991,
Machine Learning.
[4]
Ethem Alpaydin,et al.
Introduction to machine learning
,
2004,
Adaptive computation and machine learning.
[5]
Omar Adwan,et al.
Utilizing an Enhanced Cellular Automata Model for Data Mining
,
2013
.
[6]
Dipanwita Roy Chowdhury,et al.
On the Security of Designing a Cellular Automata Based Stream Cipher
,
2017,
ACISP.
[7]
S. Wolfram,et al.
Two-dimensional cellular automata
,
1985
.
[8]
Tom Fawcett,et al.
Data mining with cellular automata
,
2008,
SKDD.