Question

我尝试运行这个示例，它使用ANN Toolbox for Scilab https://burubaxair.wordpress.com/2014/03/12/artificial-neural-networks-in-scilab/

这是代码：

    T = [
1 1 1 1 1
0 0 1 0 0
0 0 1 0 0
0 0 1 0 0
0 0 1 0 0
0 0 1 0 0
0 0 1 0 0
]';
U = [
1 0 0 0 1
1 0 0 0 1
1 0 0 0 1
1 0 0 0 1
1 0 0 0 1
1 0 0 0 1
0 1 1 1 0
]';
N = [35 10 2];
W = ann_FF_init(N);
x = [1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
0, 1, 1, 1, 0]';
t_t = [1 0]';
t_u = [0 1]';
t = [t_t, t_u];
lp = [0.01, 1e-4];
epochs = 3000;
W = ann_FF_Std_batch(x,t,N,W,lp,epochs);
y = ann_FF_run(x,N,W)
disp(y)

但我收到一个错误：

-->exec('D:\Учёба\Задачи\Recognition.sce', -1)
 !--error 15 
Подматрица задана некорректно (Submatrix is incorrect).
at line      37 of function ann_FF_grad_BP called by :  
at line      25 of function ann_FF_Std_batch called by :  
W = ann_FF_Std_batch(x,t,N,W,lp,epochs);
at line      33 of exec file called by :    
exec('D:\Учёба\Задачи\Recognition.sce', -1)

错误可能在T和U矩阵中，但我不明白为什么。你能说出我做错了什么吗？谢谢！

Answer 1

您的代码中出现了2个错误：

你不应该混合测试和训练集。
测试输入必须是单列。

您的第一个错误是x = [ 1.... ]，因为它包含一个图像，而您在N中指定您有两个输出神经元。如示例中所述，您应该x = [T,U];

您的第二个错误是将x作为ann_FF_run的测试。此函数将测试输入作为单个列。但是，因为你在用5x7矩阵之前用x训练你的NN。只需将其更改为列向量。

这是经过更正和评论的代码：

T = [...
1 1 1 1 1 ...
0 0 1 0 0 ...
0 0 1 0 0 ...
0 0 1 0 0 ...
0 0 1 0 0 ...
0 0 1 0 0 ...
0 0 1 0 0 ...
]';
U = [...
1 0 0 0 1 ...
1 0 0 0 1 ...
1 0 0 0 1 ...
1 0 0 0 1 ...
1 0 0 0 1 ...
1 0 0 0 1 ...
0 1 1 1 0 ...
]';
// setting the traing set of two image
xtrain = [T,U];
// so each image as 35 pixels so N(1) is 35
// and we have two images so N($) is 2
N = [35 10 2];
// training the NN
W = ann_FF_init(N);
// The expected response for T : 1 for T, 0 for U
t_t = [1 0]';
// The expected response for T : 1 for T, 0 for U
t_u = [0 1]';
// the overall response
t = [t_t, t_u];
// some parameters
lp = [0.01, 1e-4];
epochs = 3000;
// getting the weight of the trained NN
W = ann_FF_Std_batch(xtrain,t,N,W,lp,epochs);

// testing the traing set. 
y = ann_FF_run(xtrain,N,W)
disp('Testing the traing set')
disp(y) //should get something close to t ~ [1 0 ; 0 1]

// testing a distord U
xtest1 = matrix([1, 0, 0, 0, 1;
1, 1, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
1, 0, 0, 0, 1;
0, 1, 1, 1, 1]',-1,1);
y = ann_FF_run(xtest1,N,W)
disp('Testing a distored U')
disp(y) //should get something close to t_u ~ [0 1]

//testing something different from T and U. should get nothing
xtest2 = matrix([1, 0, 0, 0, 1;
1, 1, 0, 0, 1;
1, 0, 1, 0, 1;
1, 0, 1, 0, 1;
0, 0, 1, 1, 1;
0, 0, 1, 0, 1;
0, 1, 1, 1, 1]',-1,1);
y = ann_FF_run(xtest2,N,W)
disp('Testing something neither T nor U')
disp(y)

和scilab控制台的输出

 Testing the traing set   

    0.8538757    0.1075397  
    0.1393287    0.8957439  

 Testing a distored U   

    0.1078667  
    0.9007755  

 Testing something neither T nor U   

    0.3433933  
    0.6306797

Scilab中的神经网络识别

1 个答案: