论文部分内容阅读
把二进神经网络学习算法推广至一般情形,利用汉明球及立方体的空间覆盖生成隐层神经元并对空间集合的相交、汉明球与低维空间的笛卡尔积在神经网络中的表现形式进行了深入探讨,得出几个旨在提高学习效率和减少布尔函数实现复杂性的有用结论,并融合形成完整的学习算法。
The binary neural network learning algorithm is extended to the general case, the hidden layer neurons are generated by the space covering of Hamming sphere and the cube, and the intersection of space sets is obtained. The representation of the Hamming sphere and the Cartesian product of low dimension space in the neural network Form, we come to several useful conclusions to improve the learning efficiency and reduce the complexity of the Boolean function, and merge to form a complete learning algorithm.