A Modified Differential Evolution Algorithm trained Pi-Sigma Neural Network for Pattern Classification
Sibarama Panigrahi1, Ashok Kumar Bhoi2, Yasobanta Karali3
1Sibarama Panigrahi, Department of Computer Science and Engineering, MIRC Lab, MITS Engineering College, Rayagada, Odisha, India.
2Ashok Kumar Bhoi, Department of Computer Science and Engineering, VSSUT, Burla, Sambalpur, India.
3Yasobanta Karali, Department of Computer Science and Engineering, VSSUT, Burla, Sambalpur, India.

Manuscript received on October 26, 2013. | Revised Manuscript received on November 02, 2013. | Manuscript published on November 05, 2013. | PP: 133-136 | Volume-3 Issue-5, November 2013. | Retrieval Number: E1926113513/2013©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: In this paper a modified differential evolution (DE) algorithm trained Pi-Sigma network (PSN) is used for classification. The used DE algorithm is a modification of traditional DE/rand/1/bin algorithm and novel mutation as well as crossover strategies are followed considering both exploration and exploitation. The performance of proposed methodology for pattern classification is evaluated through three well-known real world classification problems from UCI machine learning data library. The results obtained from the proposed method for classification is compared with results obtained by applying the two most popular variants of differential evolution algorithm (DE/rand/1/bin and DE/best/1/bin) and Chemical Reaction Optimization (CRO) algorithm. It is observed that the proposed method provides better classification accuracy than that of other methods.
Keywords: Differential Evolution, Higher Order Neural Network, Pi-Sigma Network, Classification.