The Judgment for Hamming Sphere Dimple in Binary Neural Networks and Its Logical Meaning
-
摘要: 在布尔空间中,汉明球突表达了一类结构清晰的布尔函数, 由于其特殊的几何特性,存在线性可分与线性不可分两种空间结构. 剖析汉明球突的逻辑意义对二进神经网络的规则提取十分重要, 然而,从线性可分的汉明球突中提取具有清晰逻辑意义的规则, 以及如何判定非线性可分的汉明球突,并得到其逻辑意义,仍然是二进神经网络研究中尚未很好解决的问题. 为此,本文首先根据汉明球突在汉明图上的几何特性, 采用真节点加权高度排序的方法, 提出对于任意布尔函数是否为汉明球突的判定算法;然后, 在此基础上利用已知结构的逻辑意义, 将汉明球突分解为若干个已知结构的并集,从而得到汉明球突的逻辑意义; 最后,通过实例说明判定任意布尔函数是否为汉明球突的过程, 并相应得到汉明球突的逻辑表达.Abstract: In boolean space, Hamming sphere dimple is able to express a kind of boolean function with clear structure. Hamming sphere dimple contains linearly separable and nonlinearly separable structures because of its special geometric characteristic. It is very important to analyze the logical meaning of Hamming sphere dimple for extracting rules from binary neural networks. However, how to extract the rules with explicit logical meaning of the linearly separable and nonlinearly separable Hamming sphere dimples and how to judge whether a nonlinearly separable boolean function is a Hamming sphere dimple have not yet been settled. To solve these problems, we firstly analyze the features of Hamming sphere dimple with Hamming-graph, and then propose an algorithm for judging whether a boolean function is linearly or nonlinearly separable Hamming sphere dimple by sorting the weighted height of the true nodes. Furthermore, we decompose Hamming sphere dimple into two known structures to obtain the logical meaning of Hamming sphere dimple by using the logical meaning of the known structures. Finally, we explain whether an arbitrary boolean function is a Hamming sphere dimple through examples, and obtain the logical meaning of the corresponding Hamming sphere dimple.
-
Key words:
- Binary neural networks /
- Hamming sphere /
- Hamming sphere dimple /
- Cartesian sphere /
- rule extraction
-
[1] Lu Yang, Wei Zhen, Gao Jun, Han Jiang-Hong. Logical meaning of Hamming sphere and its general judgement method in binary neural networks. Journal of Computer Research and Development, 2002, 39(1): 79-86(陆阳, 魏臻, 高隽, 韩江洪. 二进神经网络中汉明球的逻辑意义及一般判别方法. 计算机研究与发展, 2002, 39(1): 79-86)[2] Huynh T Q, Reggia J A. Guiding hidden layer representations for improved rule extraction from neural networks. IEEE Transactions on Neural Networks, 2011, 22(2): 264-275[3] Diederich J, Tickle A B, Geva S. Quo Vadis. Reliable and practical rule extraction from neural networks. In: Proceedings of the Advances in Machine Learning I, Studies in Computational Intelligence. Berlin, Heidelberg: Spring, 2010, 262: 479-490[4] Zhou Zhi-Hua, Chen Shi-Fu. Rule extraction from neural networks. Journal of Computer Research and Development, 2002, 39(4): 398-405(周志华, 陈世福. 神经网络规则抽取. 计算机研究与发展, 2002, 39(4): 398-405)[5] Zhou Zhi-Hua, He Jia-Zhou, Yin Xu-Ri, Chen Zhao-Qian. A statistics-based approach for rule extraction from neural networks. Journal of Software, 2001, 12(2): 263-269 (周志华, 何佳州, 尹旭日, 陈兆乾. 一种基于统计的神经网络规则抽取方法. 软件学报, 2001, 12(2): 263-269)[6] Chen Guo. Method for rule extraction from neural networks based on functional point of view. Pattern Recognition and Artificial Intelligence, 2008, 21(6): 787-793 (陈果. 一种基于功能性观点的神经网络规则提取方法. 模式识别与人工智能, 2008, 21(6): 787-793)[7] Setiono R. Extracting M-of-N rules from trained neural networks. IEEE Transactions on Neural Networks, 2000, 11(2): 512-519[8] Bader S, Holldobler S, Mayer-Eichberger B. Extracting propositional rules from feed-forward neural networks: a new decompositional approach. In: Proceedings of the 3rd International Workshop on Neural-Symbolic Learning and Reasoning. Hyderabad, India: IJCAI, 2007. 1-6[9] Bader S. Extracting propositional rules from feed-forward neural networks by means of binary decision diagrams. In: Proceedings of the 5th International Workshop on Neural-Symbolic Learning and Reasoning. Pasadena, California, USA, 2009. 22-27[10] Darbari A. Rule Extraction from Trained ANN: a Survey. Technical Report, Institute of Artificial Intelligence, Department of Computer Science, TU Dresden, Germany, 2001[11] Tsopze N, Mephu-Nguifo E, Tindo G. Towards a generalization of decompositional approach of rule extraction from multilayer artificial neural network. In: Proceedings of the 2011 International Joint Conference on Neural Networks. San Jose, California, USA: IEEE, 2011. 1562-1569[12] Lu Yang, Han Jiang-Hong, Zhang Wei-Yong. Logical relation determination criteria and equivalence rule extraction on binary neural networks. Pattern Recognition and Artificial Intelligence, 2001, 14(2): 171-176 (陆阳, 韩江洪, 张维勇. 二进神经网络逻辑关系判据及等价规则提取. 模式识别与人工智能, 2001, 14(2): 171-176)[13] Lu Yang, Han Jiang-Hong, Wei Zhen. A general judging and constructing method of SP functions in binary neural networks. Acta Automatica Sinica, 2003, 29(2): 234-241(陆阳, 韩江洪, 魏臻. 二进神经网络中SP函数的一般判别和构造方法. 自动化学报, 2003, 29(2): 234-241)[14] Lu Yang, Han Jiang-Hong, Zhang Wei-Yong. Study of Cartesian sphere in binary neural networks. Pattern Recognition and Artificial Intelligence, 2004, 17(3): 368-373 (陆阳, 韩江洪, 张维勇. 二进神经网络中笛卡尔球的研究. 模式识别与人工智能, 2004, 17(3): 368-373)[15] Yang Juan, Lu Yang, Huang Zhen-Jin, Wang Qiang. Hamming sphere dimple in binary neural networks and its linear separability. Acta Automatica Sinica, 2011, 37(6): 737-745 (杨娟, 陆阳, 黄振谨, 王强. 二进神经网络中的汉明球突及其线性可分性. 自动化学报, 2011, 37(6): 737-745)[16] Kim J H, Park S K. The geometrical learning of binary neural networks. IEEE Transactions on Neural Networks, 1995, 6(1): 237-247[17] Ma Xiao-Min, Yang Yi-Xian, Zhang Zhao-Zhi. A new framework and some results for threshold function. Chinese Journal of Computers, 2000, 23(3): 225-230 (马晓敏, 杨义先, 章照止. 一种新的阈函数的分析框架及有关结论. 计算机学报, 2000, 23(3): 225-230)[18] Hua Qiang, Zheng Qi-Lun. The Hamming-graph learning algorithm of BNN. Chinese Journal of Computers, 2001, 24(11): 1250-1255 (华强, 郑启伦. 二进神经网络的汉明图学习算法. 计算机学报, 2001, 24(11): 1250-1255)[19] Gray D L, Michel A N. A training algorithm for binary feedforward neural networks. IEEE Transactions on Neural Networks, 1992, 3(2): 176-194[20] Ma Xiao-Min, Yang Yi-Xian, Zhang Zhao-Zhi. Research on the learning algorithm of binary neural network. Chinese Journal of Computers, 1999, 22(9): 931-935 (马晓敏, 杨义先, 章照止. 二进神经网络学习算法研究. 计算机学报, 1999, 22(9): 931-935)[21] Chen F Y, Chen G R, He Q B, He G L, Xu X B. Universal perceptron and DNA-like learning algorithm for binary neural networks: non-LSBF implementation. IEEE Transactions on Neural Networks, 2009, 20(8): 1293-1301
点击查看大图
计量
- 文章访问数: 1684
- HTML全文浏览量: 39
- PDF下载量: 660
- 被引次数: 0