线性可分结构系对二进神经元的覆盖问题

Cover Problem of Linear Separable Structures to Binary Neurons

二进神经网络中每个二进神经元等价于一个线性可分函数,但每个二进神经元所表达的线性可分函数的逻辑意义仍不完全清楚。对此,首先分析了已有的几种线性可分结构系;其次,讨论了其是否覆盖了所有的二进神经元;最后,指出阈值在某些范围内二进神经元所对应的线性可分函数的逻辑意义仍不清楚,这为进一步完善二进神经元的覆盖问题指明了方向。

In binary neural networks, every neuron is equivalent to a linear separable function, however, the logical meaning of every linear separable function which is expressed by the binary neuron is still not clear. This paper firstly analysed the known linear separable structures, then discussed whether these known structures cover the whole binary neurons, finally pointed out when the threshold value is in some range, the logical meaning of the linear separable func- tion is still unclear. This result provides the way of the cover problem of binary neurons.