BP 算法的可视化研究与 BP 智能算法

A VISUAL RESEARCH INTO BP NEURAL NETWORK AND BP INTELLIGENCE ALGORITHMS

本文利用算法的可视化技术，从ＢＰ算法的基本原理出发，分析、研究了梯度、共轭自适应算法的长与短，提出了共轭梯度自适应算法的设想，并在此基础上，利用模拟退火算法思路，构制了ＢＰ智能算法。最后对实际资料进行了多种算法的对比试算，证实了共轭梯度法和ＢＰ智能算法与原算法相比，提高了速度，增强了算法对不同对象的自适应性和智能性。

In this paper, in the view of BP neural network basic principles, the authors research into the strong and weak points of the conjugate gradient self adapting alaorithm, propose the thought of the conjugate gradient self adapting algorithm and based on the thought, innovate on BP intelligence algorithms by simulating the annealing thought. By comparing, testing and computing different algorithms on the same samples of different facies belts, the authors prove that conjugate gradient self adapting and BP intelligence algorithms have improved the speed and enchanced their self adaptability and intelligence to different objects comparing with the original algorithms.