| 摘要: |
| 扩展互信息分离算法采用单隐层神经网络近似算法代价函数中的非线性函数，可调节
的参数有限，需要多次迭代才能收敛，从而导致收敛速度较慢。针对这一问题，
采用双隐层神经网络近似非线性函数，以分离结果的互信息最小化作为代价函数，采用梯度
下降方法对代价函数进行优化，增加了可调节参数数量。仿真实验结果表明，改进后的算法
相对原算法收敛速度更快，误差更小。 |
| 关键词: 非线性独立分量分析 扩展互信息分离算法 多层感知机 双隐层神经网络 |
| DOI： |
|
| 基金项目:吉林省科技发展计划项目(201101110) |
|
| Extended mutual information separation algorithms based on multi-hidden layer |
| ZHAO Li-quan,CAI Bang-gui |
| () |
| Abstract: |
| Extended mutual information separation(EMISEP) algorithm uses a single hidden
layer neural network to approximate nonlinear function of cost function, so the
adjustable parameter is limited and it needs more iteration times to converge, w
hich leads to relatively slow convergence speed. To overcome this pr
oblem, this paper uses double hidden layer perceptions to approximate nonlinear
function of cost function, and uses mutual information minimum of separation sig
nals as cost function, which is optimized by gradient descent method. This incre
ases the number of adjustable parameters. The simulation results prove that the imp
roved algorithm has faster convergence speed and smaller error comparing with t
he original algorithm. |
| Key words: nonlinear independent component analysis extended mutual information separation
algorithm multilayer perception double hidden layer neural network |
