Selection of the number of hidden neurons in MLP

Asked Aug 15 '22 at 21:09

Active Aug 18 '22 at 12:18

Viewed 254 times

I am trying to do MLP regressor for my data (6 INPUTS & 19 OUTPUTS). I am using 2 hidden layers MLP. I have 100 datapoints.

I am following the guidelines mentioned in this post for the selection of the number of my hidden nodes.

According to the above source, the number of my hidden neurons should be between 6 and 15. So I've done various trial and error but I AM STILL GETTING A VERY LOW r^2 after CV.

However, when I select the number of my hidden neurons to be (200,200) I get a very good r^2 after CV. What's the logic behind that? Thank you.

edited Aug 18 '22 at 12:18

Scriddie

2,244
6
13

asked Aug 15 '22 at 21:09

Z47

Is the $R^2$ value measured on the training data or the testing data? – Sycorax Aug 16 '22 at 00:47
@Sycorax testing data. – Z47 Aug 16 '22 at 00:49

Selection of the number of hidden neurons in MLP

0 Answers0