Journal of KIISE:Software and Applications (한국정보과학회논문지:소프트웨어및응용)

Korean Institute of Information Scientists and Engineers (한국정보과학회)
권호 표지

Improvement of Classification Rate of Handwritten Digits by Combining Multiple Dynamic Topology-Preserving Self-Organizing Maps

다중 동적 위상보존 자기구성 지도의 결합을 통한 필기숫자 데이타의 분류율 향상

  • 김현돈 (연세대학교 컴퓨터과학과) ;
  • 조성배 (연세대학교 컴퓨터과학과)
  • Published : 2001.12.01
Download PDF
⟨ Previous Next ⟩

Abstract

Although the self organizing map (SOM) is widely utilized in such fields of data visualization and topology preserving mapping, since it should have the topology fixed before trained, it has some shortcomings that it is difficult to apply it to practical problems, and classification capability is quite low despite better clustering performance. To overcome these points this paper proposes the dynamic topology preserving self-organizing map(DTSOM) that dynamically splits the output nodes on the map and trains them, and attempts to improve the classification capability by combining multiple DTSOMs K-Winner method has been applied to combine DTSOMs which produces K outputs with winner node selection method. This produces even better performance than the conventional combining methods such as majority voting weighting, BKS Bayesian, Borda, Condorect and reliability sum. DTSOM remedies the shortcoming of determining the topology in advance, and the classification rate increases significantly by combing multiple maps trained with different features. Experimental results with handwritten digit recognition indicate that the proposed method works out to problems of conventional SOM effectively so to improve the classification rate to 98.1%.

자기 구성지도는 데이타 시각화, 위상보존 매핑 등의 분야에서 널리 사용되고 있지만. 학습이 되기 전에 위상을 미리 고정시켜야 하기 때문에 실제 문제에 적용하기 어렵다는 것과 클러스터링 능력에 비해 분류율이 낮다는 결점이 있다. 이를 해결하기 위해서 자기구성 지도의 출력 노드를 동적으로 분화하고 분화된 노드를 파습하는 동적 위상보존 사기구성 지도를 제안하고, 이를 다중 결합함으로써 분류율을 향상 시켰다. 동적 위상보존 자기구성 지도의 결함 방법으로는 자기구성 지도의 K개 노드가 출격을 내도록하는 K-Winner 방법 및 K-Winner+ 가중치 방법이 제안되었는데, 이는 다수결 투표, 가중치, BKS, Byayesian, Borda, Condorect, 신뢰값 합산 등의 기존 결합 방법보다도 우수한 결과를 나타내었다. 동적 위상보존 자기 구성 지도를 통해서 위상을 고정 시켜야 하는 결점을 해결할 수 있었고. 서로 다른 특징으로 학습된 동적 위상보존 자기구성 지도들을 결합하여 분류 능력을 향상시킬 수 있었다. 필기 숫자데이타로. 실험한 결과, 제안한 방법이 자기구성 지도의 결점을 효과적으로 해결하여 98.1% 의 높은 인식률을 보였다.

Keywords

References

  1. B. Fritzke, 'Growing grid - A self-organizing netwroks with constant neighborhood range and adaptation strength.' Neural Processing Letters. Vol. 2. No. 5. 9-13, 1995 https://doi.org/10.1007/BF02332159
  2. B. Fritzkc, 'Growing self-organizing netwroks - Why?,' ESANN'96, p.61-72. 1996
  3. T. Kohonen, 'Self-organized formation of topologically correct feature maps,' Biol. Cyb., 43:59-69, 1982 https://doi.org/10.1007/BF00337288
  4. T. Kohonen, Self-Organizing Maps, Springer, Berlin Heidelberg, 1995
  5. K. Kiviluoto, 'Topology Preservation in Self-Organizing Maps,' IEEE Transactions on Neural Networks, 1996
  6. T. Villmann, R Der, M. Herrmann, and T. M. Martinetz, 'Topology Preservation in Self-Organizing Feature Maps : Exact Definition and Measurement,' IEEE Transactions on Neural Networks, vol 8, no. 2, 1997 https://doi.org/10.1109/72.557663
  7. A. Konig, 'Interactive Visualization and Analysis of Hierarchical Neural Projections for Data Mining,' IEEE Transactions on Neural Networks. vol 11, no. 3, 2000 https://doi.org/10.1109/72.846733
  8. 백종현， '다중 인식기의 다단계 결합을 통한 무제약 필기숫자 인식'. 연세대학교 대학원, 1996
  9. L. Lam, C. Y. Suen. 'A Theoritical analysis of the application of majority voting to pattern recognition,' Proceeding of the 12th ICPR, pp. 418-420, 1994 https://doi.org/10.1109/ICPR.1994.576970
  10. L. Xu, A. Krzyzak, C. Y. Sucn, 'Methods of combining multiple classifier and their applications to handwriting recognition,' IEEE Transactions on Systems, vol. SMC-22, no 3, pp. 418-435, 1992 https://doi.org/10.1109/21.155943
  11. C. Y. Suen, Y. S. Huang, 'Multi-Expert Systems for pattern Recognition,' Proceedings of 2nd pacific Rim International Conference on Artificial Intelligence, pp. 15-20, 1992
  12. T. K. Ho, 'On multiple classifier systems for pattern recognition,' Proceeding of 11th ICPR, vol.1. pp, 84-87, 1992 https://doi.org/10.1109/ICPR.1992.201727
  13. S. B. Cho, 'Self-organizing map with dynamical node splitting: Application to handwritten digit recognition,' Neural Computation, Vol. 9. 1345-1355, 1997 https://doi.org/10.1162/neco.1997.9.6.1345
  14. J. C. Bioch, O. V. D. Meer, R Potharst, 'Classification using bayesian neural nets,' IEEE International Conference on Neural Networks, vol, 3, pp. 1488-1493, 1996 https://doi.org/10.1109/ICNN.1996.549120
  15. A. Khotanzad, and C. Chung, 'Hand written digit recognition using BKS combination of neural network classifiers,' Proceedings of the IEEE Southwest Symposium on Image Analysis and interpretation, pp. 94-99, 1994 https://doi.org/10.1109/IAI.1998.666880