使用分段逼近(piecewise approximation)算法计算超越函数,以神经网络中最常用的双曲正切型(tanh)传输函数为例来分析逼近精度同分段数、有限字长之间的关系。 -Using segmented approximation (piecewise approximation) algorithm for computing transcendental functions, to the most commonly used neural network type hyper

该程序是分段函数： akm(m,n){ n+1 m =0 akm(m-1,1) m!=0,n=0 akm(m-1),(akm(m,n-1)) m!=0,n!=0 实现递归和非递归的算法 非递归算法可能使计算时间减少-The program is piecewise function: Akm (m, n) { N+ 1 m = 0 Akm (m-1, 1) m! = 0, n = 0 Akm (m-1) (akm (m, n-1)) m! = 0, n.=