In this demo, I use the EM algorithm with a Rauch-Tung-Striebel smoother and an M step, which I ve recently derived, to train a two-layer perceptron, so as to classify medical data (kindly provided by Steve Roberts and Will Penny from EE, Imperial College). The data and simulations are described in: Nando de Freitas, Mahesan Niranjan and Andrew Gee Nonlinear State Space Estimation with Neural Networks and the EM algorithm After downloading the file, type "tar -xf EMdemo.tar" to uncompress it. This creates the directory EMdemo containing the required m files. Go to this directory, load matlab5 and type "EMtremor". The figures will then show you the simulation results, including ROC curves, likelihood plots, decision boundaries with error bars, etc. WARNING: Do make sure that you monitor the log-likelihood and check that it is increasing. Due to numerical errors, it might show glitches for some data sets.
標(biāo)簽：
Rauch-Tung-Striebel
algorithm
smoother
which
上傳時(shí)間：
2016-04-15
上傳用戶：zhenyushaw
% 生成訓(xùn)練樣本集
clear all;
clc;
P=[110 0.807 240 0.2 15 1 18 2 1.5;
110 2.865 240 0.1 15 2 12 1 2;
110 2.59 240 0.1 12 4 24 1 1.5;
220 0.6 240 0.3 12 3 18 2 1;
220 3 240 0.3 25 3 21 1 1.5;
110 1.562 240 0.3 15 3 18 1 1.5;
110 0.547 240 0.3 15 1 9 2 1.5];
0 1.318 300 0.1 15 2 18 1 2];
T=[54248 162787 168380 314797;
28614 63958 69637 82898;
86002 402710 644415 328084;
230802 445102 362823 335913;
60257 127892 76753 73541;
34615 93532 80762 110049;
56783 172907 164548 144040];
@907 117437 120368 130179];
m=max(max(P));
n=max(max(T));
P=P'/m;
T=T'/n;
%-------------------------------------------------------------------------%
pr(1:9,1)=0; %輸入矢量的取值范圍矩陣
pr(1:9,2)=1;
bpnet=newff(pr,[12 4],{'logsig', 'logsig'}, 'traingdx', 'learngdm');
%建立BP神經(jīng)網(wǎng)絡(luò)， 12個(gè)隱層神經(jīng)元，4個(gè)輸出神經(jīng)元
%tranferFcn屬性 'logsig' 隱層采用Sigmoid傳輸函數(shù)
%tranferFcn屬性 'logsig' 輸出層采用Sigmoid傳輸函數(shù)
%trainFcn屬性 'traingdx' 自適應(yīng)調(diào)整學(xué)習(xí)速率附加動(dòng)量因子梯度下降反向傳播算法訓(xùn)練函數(shù)
%learn屬性 'learngdm' 附加動(dòng)量因子的梯度下降學(xué)習(xí)函數(shù)
net.trainParam.epochs=1000;%允許最大訓(xùn)練步數(shù)2000步
net.trainParam.goal=0.001; %訓(xùn)練目標(biāo)最小誤差0.001
net.trainParam.show=10; %每間隔100步顯示一次訓(xùn)練結(jié)果
net.trainParam.lr=0.05; %學(xué)習(xí)速率0.05
bpnet=train(bpnet,P,T);
%-------------------------------------------------------------------------
p=[110 1.318 300 0.1 15 2 18 1 2];
p=p'/m;
r=sim(bpnet,p);
R=r'*n;
display(R);
標(biāo)簽：
2013
算法
工具箱
上傳時(shí)間：
2016-05-28
上傳用戶：shanqiu
