论文部分内容阅读
在基于FPGA的神经网络设计中,提出一种采用直接坐标旋转数字计算机(CORDIC)算法计算神经元激励函数ex的方法,依靠移位和求和能够实现快速、精确的指数函数计算,较查表法和间接CORDIC算法既节省了大量片内资源,又提高了计算速度和精度.利用Xilinx公司ISE开发工具进行仿真实验,结果表明直接CORDIC算法的计算速度是间接CORDIC算法的14倍,证明了该算法计算指数函数的快速性与精确性.
In the design of neural network based on FPGA, a method of computing the exponential function of neuron using the CORDIC algorithm is proposed, which can quickly and accurately calculate the exponential function by shifting and summing. The method and indirect CORDIC algorithm not only save a large amount of on-chip resources, but also improve the speed and accuracy of computation.Using Xilinx ISE development tools to simulate the experiment results show that the direct CORDIC algorithm is 14 times faster than the indirect CORDIC algorithm, The algorithm calculates the speed and accuracy of exponential functions.