How to Decide size of Neural Network like number of neurons in a hidden layer & Number of hidden layers?

Question

sunny B 2013-4-18

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/72654-how-to-decide-size-of-neural-network-like-number-of-neurons-in-a-hidden-layer-number-of-hidden-lay

评论： Jeong_evolution 2016-12-20

Hi friends, I want to design a neural network which should give one output with five inputs and i have input samples are 432. So please suggest how to design neural network and which type of neural network i should and how to decide number of hidden layers and no of neurons in each hidden layer.

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

Greg Heath 2013-4-19

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/72654-how-to-decide-size-of-neural-network-like-number-of-neurons-in-a-hidden-layer-number-of-hidden-lay#answer_82772

Use 1 hidden layer.

Use fitnet for curve-fitting or regression.

Use patternnet for pattern-recognition or classification. For c categories with indices 1:c, the target matrix should contain columns of eye(c). The relationship between target columns and class indices is target = ind2vec(classindices) and classindices = vec2ind(target). Since since the sum of the target columns is 1, one of the rows (usually the last) can be omitted. This reduction is used mostly when c=2.

Find a good value for the number of hidden nodes, H, by trial and error. Create numH*Ntrials nets in a double loop: An outer loop H = Hmin:dH:Hmax over number of hidden nodes and an inner loop i = 1: Ntrials over number of random trn/val/tst data divisions and random weight initialization trials for each value of H.

I typically use numH ~ 10, Ntrials ~ 10 and tabulate results in Ntrials X numH matrices.

Many examples can be obtained by searching the NEWSGROUP and ANSWERS using the keywords

Greg Ntrials Nw

Hope this helps.

Thank you for formally accepting my answer

Greg