Parallel implementation of handwritten digit recognition system using self-organizing map
WANG Yi-mu1, PAN Yun1, LONG Yan-chen1, YAN Xiao-lang1, HUAN Ruo-hong2
1. Institute of VLSI Design, Zhejiang University, Hangzhou 310027, China; 2. College of Computer Science and
Technology, Zhejiang University of Technology, Hangzhou 310023, China
A simplified self-organizing map (SOM) algorithm and its parallel hardware architecture were proposed in order to handle the complex problem of hardware implementation of SOM. The weight update function of the conventional SOM contains multiplication, square and exponential operations, which makes parallel realization difficult. Traditional simplified SOM methods have low accuracy on classification, especially when encountering complex applications such as handwritten digit recognition. The proposed SOM algorithm has reduced all complex computations without sacrificing accuracy, and full parallel hardware implementation is possible. The proposed hardware system can proceed at 50 MHz and achieve a performance of 400 MCUPS, which means that learning a single input pattern will take only 200 ns of time. When applying the handwritten digit recognition on the MNIST database, the proposed system can recognize over 81% of the patterns correctly, which is almost the same accuracy as the conventional SOM, but is much better than other simplified SOM methods.
WANG Yi-mu, PAN Yun, LONG Yan-chen, YAN Xiao-lang, HUAN Ruo-hong. Parallel implementation of handwritten digit recognition system using self-organizing map. JOURNAL OF ZHEJIANG UNIVERSITY (ENGINEERING SCIENCE), 2014, 48(4): 742-747.
［1］ TEUVO K. The self-organizing map ［C］∥ Proceedings of IEEE. New York: IEEE, 1990, 78(1): 1464-1480.
［2］ 颜学峰, 陈德钊,胡上序.复杂模式保留拓扑的平面映射及其应用［J］. 浙江大学学报：工学版, 2001, 35(5): 529-533.
YAN Xue-feng, CHEN De-zhao, HU Shang-xu. Topology-preserving map of complex patterns group and its application ［J］. Journal of Zhejiang University: Engineering Science, 2001, 35(5): 529-533.
［3］ TEUVO K, SAMUEL K, KRISTA L, et al. Self organization of a massive document collection ［J］. IEEE Transactions on Neural Networks, 2000, 11(3): 574-585.
［4］ SILVEN O, NISKANEN M, KAUPPINEN H. Wood inspection with non-supervised clustering ［J］. Machine Vision and Applications, 2003, 13(5/6): 275-285.
［5］ MUTHURAMALINGAM A, HIMAVATHI S, SRINIVASAN E. Neural network implementation using FPGA: issues and application ［J］. International Journal of Information Technology, 2008, 4(2): 86-92.
［6］ RUPING S, RUCKERT U, GOSER K. Hardware design for self organizing feature maps with binary input vectors ［C］∥ Proceedings of the International Workshop on Artificial Neural Networks. Sitges: Springer, 1993: 488-493.
［7］ JORGE P, MAURICIO V, ANDRES V. Digital hardware architectures of Kohonen’s self organizing feature maps with exponential neighboring function ［C］∥ IEEE International Conference on Reconfigurable Computing and FPGA. Mexico: IEEE, 2006: 1-8.
［8］ YAMAKAWA T, HORIO K, HIRATSUKA T. Advanced self organizing maps using binary weight vector and its digital hardware design ［C］∥ Proceedings of the 9th International Conference on Neural Information. Singapore: IEEE, 2002: 1330-1335.
［9］ APPIAH K, HUNTER A, MENG H Y, et al. A binary self-organizing map and its FPGA implementation ［C］∥ Proceedings of International Joint Conference on Neural Networks. Atlanta: IEEE, 2009: 164-171.
［10］ OMONDI A R, RAJAPAKSE J C. FPGA implementations of neural networks ［M］. Netherlands: Springer, 2006.
［11］ AGUNDIS R, GIRONES G, PALERO C, et al. A mixed hardware software SOFM training system ［J］. Computacióny Sistemas, 2008, 4: 349-356.
［12］ LECUN Y, CORTES C. The MNIST database of handwritten digits ［DB/OL］. http:∥yann.lecun.com/exdb/mnist/.
［13］ APPIAH K, HUNTER A, DICKINSON P, et al. Implementation and applications of tri-state self-organizing maps on FPGA ［J］. IEEE Transactions on Circuits and Systems for Video Technology, 2012, 22(8): 1150-1160.
