Contents lists available at ScienceDirect Biologically Inspired Cognitive Architectures journal homepage: www.elsevier.com/locate/bica Research article Biologically inspired cellular automata learning and prediction model for handwritten pattern recognition Aamir Wali ⁎ , Mehreen Saeed Department of Computer Science, FAST-NUCES, Faisal Town, Lahore, Pakistan ARTICLE INFO Keywords: Cellular automata CALP Conway’s game of life Handwritten pattern recognition Data generation and over-sampling techniques Neural ensemble Ensemble methods ABSTRACT In this study, we propose an ensemble learning architecture called “Cellular Automata Learning and Prediction” (CALP) model, for classification of handwritten patterns. We further propose that every handwritten pattern is an array of living cells or organisms that both interact and are affected by one another. Since the cells impact one another, and have the ability to die and reproduce, we extend this analogy to growth and evolution. Thus every pattern can grow and evolve. We use cellular automata (CA) to model this behavior as it has been used as a default model for various biological systems. Proposed architecture allows the handwritten patterns to evolve or grow using various parameters that control how the cells interact with each other. Then these different evolved patterns are used to train independent classifiers which are then combined together to form an ensemble. The idea is to captures more variations in a handwritten data set than the typical standalone classifiers or their ensembles. The method is applied on 5 handwritten data sets using 5 different classifiers. The experimental results show that our model obtains better classification accuracy on all 5 data sets, even on a small-sized training data. We also compare the performance of CALP with other over-sampling methods. 1. Introduction The idea of cellular automata (CA) was originally conceived by Stanislaw Ulam and John von Neumann in 1040s. However, a more systematic study on CA was done much later in 1990 and then in 2002 by Wolfram in his well known book A new kind of Science (Wolfram, 2002). Later on, many interesting properties of CA were discovered, and they were applied in different areas of computer science such as image processing (Popovici & Popovici, 2002; Rosin, 2010), machine learning (Wongthanavasu & Ponkaew, 2013), data mining (Fawcett, 2008), cryptography (Seredynski, Bouvry, & Zomaya, 2004), allocation hub location problem (Saghiri & Meybodi, 2018), etc. Recently, it has been shown that CA can be successfully applied to simulate cell and bacterial growth, and to model urban growth to predict the evolution of society and population (Abolhasani, Taleai, Karimi, & Rezaee Node, 2016; Berberoğlu, Akın, & Clarke, 2016). They have also been applied for generating test cases for software testing (Bhasin, 2014; Bhasin, Singla, & Sharma, 2013). This is because CA being a dynamic system that consists of cells, and a set of rules that explain how the cells evolve with time, gives them the ability to reproduce using self-generating patterns, and to model evolution and growth of a pattern. In this paper, we first propose that a handwritten pattern is a col- lection of living cells, and then propose one such model that exploits the important properties of interacting cells and the self generating char- acteristics of cells to evolve data. We also show how the fascinating abilities of cells modeled using CA can be amalgamated with hand- written pattern recognition system techniques to give rise to a metho- dology, which evolves the patterns using CA and use these evolved patterns for classification. The CA was used particularly because it is used in various biological systems (Akdur, 2011; Vitvitsky, 2016; Youssef, 2013). Since our pattern can evolve and grow to new patterns, this is also related to synthetic data generation. The idea of synthetic pattern generation is not uncommon in the area of classification and pattern recognition. Synthetic data set generation, besides being used for the purpose of increasing number of training samples, is also used to bal- ance imbalanced data sets (Charte, Rivera, del Jesus, & Herrera, 2015; Fernández-Navarro, Hervás-Martínez, & Gutiérrez, 2011; Sáez, Krawczyk, & Woźniak, 2016). Data generation is also of interest to the deep learning community. A major disadvantage of state-of-the-art convolutional neural network is the need for huge amount of data, whereas our proposed method is capable of performing well on small- sized data by simulating variations of existing data. The difference between CA and other data generation techniques is that the other methods take a set of samples or data points and generate new samples within them using interpolation, nearest neighbor, or https://doi.org/10.1016/j.bica.2018.04.001 Received 13 March 2018; Accepted 9 April 2018 ⁎ Corresponding author. E-mail addresses: aamir.wali@nu.edu.pk (A. Wali), mehreen.saeed@nu.edu.pk (M. Saeed). Biologically Inspired Cognitive Architectures xxx (xxxx) xxx–xxx 2212-683X/ © 2018 Elsevier B.V. All rights reserved. Please cite this article as: Wali, A., Biologically Inspired Cognitive Architectures (2018), https://doi.org/10.1016/j.bica.2018.04.001