| 摘要: |
| 许多实际的模式识别问题如对手写体汉字的识别,都属于大规模的模式识别问题.目前,传统的神经网络算法对这类问题尚无有效的解决办法.在球邻域模型的基础上提出一种可用于大规模模式识别问题的神经网络训练算法,试图加强神经网络解决大规模问题的能力,并用手写体汉字识别问题检验其效果.实验结果揭示了所提算法是解决大规模模式识别问题的一个有效且具有良好前景的方法. |
| 关键词: 神经网络 模式识别 字符识别 训练算法 球邻域模型 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金资助项目(69823001);国家重点基础研究发展规划973资助项目(G1998030509);高等学校博士学科点专项科研基金(98000335) |
|
| A Neural Network Algorithm for Large Scale Pattern Recognition Problems |
|
WU Ming rui,ZHANG Bo
|
| Abstract: |
| Many practical pattern recognition problems, such as recognition of handwritten Chinese characters belong to the pattern recognition problems of large scale. Now conventional ANN (artificial neural network) algorithms cannot solve this set of problems efficiently. In this paper, a neural network algorithm based on the sphere neighborhood model is introduced, aiming at enhancing the neural network's ability to solve the pattern recognition problems of large scale. The performance of the algorithm is tested with the handwritten Chinese character recognition problem. Experimental result show that the proposed algorithm is competent and has well prospects to this set of problems. |
| Key words: neural network pattern recognition character recognition training algorithm sphere neighborhood model |
