Implementing a Neural Network

Question

Tony Stark 2022-10-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1835348-implementing-a-neural-network

评论： Walter Roberson 2022-10-26

I am trying to create a linear neural network that takes in 11 inputs and gives out 5 outputs.

I've been using the following documentation as reference: https://www.mathworks.com/help/deeplearning/ug/linear-neural-networks.html

Now, x is a 75 by 11 array, of inputs for the training data.

Meanwhile, y is a 5 by 75 array, of true label outputs for the training data.

So far I have tried:

P = x'
T = y
net = linearlayer;
net = configure(net,P,T);
net.IW{11,5}
net.b{1} 

I am getting an error when initializing the weights, what am I doing wrong? Since there is going to be 10 inputs, I assume that there will be 11 weights. Then there should be 1 bias weight.

Overall structure is supposed to be 11 inputs for the input layer, 1 hidden layer, and an output layer of 5 outputs. (Hence, y has 5 rows. Also, hence why x has 11 columns)

I am only trying to set up the linear neural network, not at the point of training it.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2022-10-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1835348-implementing-a-neural-network#answer_1083568

在 MATLAB Online 中打开

x = rand(75,11);
y = randi(5, 5, 75);
P = x';
T = y;
net = linearlayer;
net = configure(net,P,T);
net.IW
ans = 1×1 cell array
    {5×11 double}
net.IW{1}(5,11)
ans = 0
net.b{1} 
ans = 5×1
     0
     0
     0
     0
     0

2 个评论
显示无隐藏无

Tony Stark 2022-10-26

Thank you for your answer, it is really helpful. Although, since there are 11 inputs, shouldn't there be 11 weights? Why is the answer just 0?

Walter Roberson 2022-10-26

There are 11 scalar weights for each of the 5 outputs, a 5 x 11 array. Asking to look at the 11, 5 configuration is asking to look at the initialization weight for the last step for the last output, a scalar.

请先登录，再进行评论。