Use matlab trained patternnet in C/other platform

1 次查看（过去 30 天）

显示更早的评论

hassan mahmood 2018-2-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/385104-use-matlab-trained-patternnet-in-c-other-platform

评论： hassan mahmood 2018-2-28

在 MATLAB Online 中打开

I trained a neural network for pattern recognition with 100 neurons and 33 inputs. I used the command

net = patternnet(100);

I got all the weight and other matrices. Now I want to use this trained network on ARM platform.

I have been told that if you have all the matrices you can implement the network in any language and can get approximately same results. All you need is simple feed-forward neural network for which weights have been trained.

I don't know the maths of NN. Please guide me how to implement feed-forward network by myself.

How do I know that how many neurons are connected to how many inputs in how many manners in a feedforward network created by matlab with 100 neurons for 33 inputs.

I only know the structure of simple neural network ,which is for 3 inputs and 3 neurons are used in hidden layer. Where each neuron is connected to each input. And each neuron has activation function.and then there is output layer. I can code this network in any language.

I know its naive and basic level but I don't really know that how 100 neuron with 33 input make a feed-forward network.and further how can I know the structure of trained neural network in Matlab so that I can code the same in any other language and can use the trained weights on the same network. In short how matlab generate feed-forward neural network with 100 neurons for 33 inputs and 256 output classes. What will be the structure? So that I can implement it on other platforms.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

hassan mahmood 2018-2-28

In short how matlab generate feed-forward neural network with 100 neurons for 33 inputs and 256 output classes. What will be the structure? So that I can implement it on other platforms.

请先登录，再进行评论。

请先登录，再回答此问题。