Title :
Cellular Automata learning algorithm for classification
Author :
Wongthanavasu, Sartra ; Ponkaew, Jesada
Author_Institution :
Dept. of Comput. Sci., Khon Kaen Univ., Khon Kaen, Thailand
Abstract :
Cellular Automata (CA) were proved as universal computation model. This means that CA are capable of modeling any problems to arrive at the solutions. Since then, researchers have paid efforts to seek the possible classifiers being constructed on this model. However, there is a number of papers to date haing reported classification with limited applications using cellular automata. This paper presents a novel state of the art classification model on the basis of cellular automata. The proposed classifier was developed on the proposed Decision Support Elementary Cellular Automata (DS-ECA). It comprises double rule vectors and a decision function. Its structure comprises two layers; the first layer is employed to evolve an input pattern into feature space and the other then interprets the feature space to binary answer through the decision function. For performance evaluation, six datasets consisting of binary and non-binary features are implemented in comparison with Support Vector Machines (SVM) with the optimal kernel using k-fold cross validation. In this respect, the proposed method outperforms SVM for binary featured datasets, and reports equivalent and better results on average for non-binary features.
Keywords :
cellular automata; decision theory; feature extraction; learning (artificial intelligence); pattern classification; support vector machines; DS-ECA; SVM; binary answer; binary featured dataset; cellular automata learning algorithm; classification model; decision function; decision support elementary cellular automata; double rule vectors; feature space; input pattern evolution; k-fold cross validation; nonbinary features; optimal kernel; performance evaluation; support vector machines; universal computation model; Manganese; Polynomials; Support vector machines; Testing; Cellular automata; classification; supervised learning algorithm; support vector machines;
Conference_Titel :
Electrical Engineering Congress (iEECON), 2014 International
Conference_Location :
Chonburi
DOI :
10.1109/iEECON.2014.6925938