New evolutionary neural networks
-
摘要: 目前的进化神经网络模型大多采用遗传算法进行网络进化设计.而研究表明,这种进化神经网络存在遗传编码、遗传操作及网络结构限制等很多问题;而采用进化规划是一种很好的途径.鉴于此,为了克服传统进化规划算法的不足,结合作者提出的快速免疫进化规划提出了一种网络连接权值及其拓扑结构同时进化优化的新型进化神经网络模型.最后,通过典型的异或分类问题(XOR)比较了该模型同BP神经网络及传统进化神经网络的计算性能,发现它不但计算精度好,而且计算效率高.Abstract: Nowadays, the genetic algorithm is a main evolutionary algorithm in ev olutionary neural network study. But the previous researches show that, this kin d of evolutionary neural network model have many shortcomings, such as genetic c ode problem, genetic operation problem and restriction on structure of neural ne twork, \%et al\%. And the evolutionary programming is a good method. To overcome the shortcomings of traditional evolutionary programming, combining the immunized e volutionary programming proposed by author and BP neural network, a new evolutio nary neural network model whose architecture and connection weights evolve simul taneously was proposed. Through the typical XOR problem, the new model was compa red and analyzed with BP neural network and traditional evolutionary neural netw ork. The computing results show that the precision and efficiency of the new mod el are all good.
-
[1] 陈颖琪,余松煜,赵 弈.进化计算与人工神经网络的结合[J].红外与激光工程, 19 99, 28(4):6~9 Chen Yingqi, Yu Songyu, Zhao Yi. Combination of evolutionary computation and artificial neural network[J]. Infrared and Laser Engineering, 1999, 28 (4):6~9 (in Chinese) [2] Yao Xin.A review of artificial neural networks[J]. Int J of Intelligent Systems, 1993, 8(4):539~567 [3]Yao Xin.Evolutionary artificial neural networks[J]. Int J of Neural Systems, 1993, 4(3):203~222 [4]Muhlenbein H. Limitations of multi-layer preceptron networks-steps towards genetic neural networks[J]. Parallel Computing, 1990, 14:249~260 [5]Baluja S. Evolution of an artificial neural network based autonomous land vehicle controller[J]. IEEE Trans on SMC, 1996,26:450~463 [6]Millar G F, Todd P M, Hegde S U. Designing neural network using genetic algorithms . In:Proc 3rd Int Conf Genetic Algorithms and Their Applications . San Mateo:Morgan Kaufmann, 1989. 379~384 [7]kourikhine A N. An evolutionary algorithm for designing feed-forward neural networks . In:7th Int Conf on EP'98 . New York:Springer, 1998. 629~635 [8]Takahashi H, Agui T, Nagahshi H. Designing adaptive neural network architectures and their learning parameters using genetic algorithms .In:Proc Science of ANN . Florido, 1993. 208~215 [9]Yao Xin,Liu Yong.A new evolutionary system for evolving artificial neural networks[J]. IEEE Trans on NN, 1997, 8(3):694~713 [10] Fang Jian,Xi Yugeng. Neural network design based on evolutionary programming[J]. Artificial Intelligence in Engineering, 1997, 11:155~161 [11]Maniezzo V. Genetic evolution of the topology and weight distribution of neural networks[J]. IEEE Trans on NN, 1994, 5(1):39~53 [12]Angeline P J, Saunders G M, Pollack J B. An evolutionary algorithm that constructs recurrent neural networks[J]. IEEE Trans on NN, 1994, 5(1):54~65 [13]McDonnell J R, Waagen D. Evolving recurrent perceptrons for time-series modeling[J]. IEEE Trans on NN, 1994, 5(1):24~38 [14] 靳 蕃. 神经计算智能基础——原理·方法[M]. 成都:西南交通大学出版社, 2000 Jin Fan. Foundation of neural computation intelligence—principles and methods[M]. Chengdu:Southwest Jiaotong University Press, 2000 (in Chinese)
点击查看大图
计量
- 文章访问数: 2394
- HTML全文浏览量: 94
- PDF下载量: 1781
- 被引次数: 0