RBF神经网络中指数函数e^x的FPGA实现

The Realization of Exponential Function ex in RBF Neural Network

RBF径向基神经网络在工程中,尤其是各种智能控制中的应用十分广泛。其隐含层的非线性激活函数经常采用高斯径向基函数,这一函数为一指数函数。指数函数用硬件实现起来比较困难。本文提供了一种应用CORDIC算法在FPGA上实现ex的方法。这种方法具有硬件实现简单、误差小和便于VLSI等优点。

RBF Radial Basis Function Neural Network is now use d very abroad in engineering, especially in various intelligent controls. And the activate function of it's hidden layer usually adopts Gauss RBF function, which is an exponential function. But the realization of exponential function is difficult. This thesis offers a method -using CORDIC arithmetic in FPGA-that can implement ex. And this method has many good characteristic: simple for hardware realization?error smallness and convenient for VLSI and so on.