CAJ | 학술논문

神经网络控制算法是近年来的研究热点，其中的BP神经网络控制算法是最常用的控制算法，它是前馈网络中的核心部分，可以无限逼近任意非线性，实现数据压缩。本文运用BP神经网络控制算法，借助Matlab神经网络工具箱函数，编写程序，首先做了一个非线性函数的逼近，其次重点运用BP神经网络实现图像压缩，通过试验比较，选择合适的训练函数，设定最小误差和最大训练步数，实现图像压缩，并计算了图像的信噪比及峰值信噪比，比较了不同压缩比下图像质量，最后分析了BP神经网络的局限性。
신경망락공제산법시근년래적연구열점，기중적BP신경망락공제산법시최상용적공제산법，타시전궤망락중적핵심부분，가이무한핍근임의비선성，실현수거압축。본문운용BP신경망락공제산법，차조Matlab신경망락공구상함수，편사정서，수선주료일개비선성함수적핍근，기차중점운용BP신경망락실현도상압축，통과시험비교，선택합괄적훈련함수，설정최소오차화최대훈련보수，실현도상압축，병계산료도상적신조비급봉치신조비，비교료불동압축비하도상질량，최후분석료BP신경망락적국한성。
In recent years, neural network has become a hot topic. BP neural network control algorithm is the most commonly used neural network control, and it is the core part of the feedforward network. It can approximate any nonlinear and can realize the data compression. This paper uses BP neural network control algorithm with the help of neural network toolbox function of Matlab, choosing the proper training function, set ing the minimum error and maximum training step, writing code, creating a three layer network, approximating a nonlinear function firstly,then implementing image compression, calculating the PSNR and SNR of the image, comparing the quality of the image under the condition of dif erent compression ratio, and analyzes the disadvantages of BP network.