2
$\begingroup$

I tried training a neural network with various numbers of hidden lawyers etc where the training set was just pairs of different side length and the answer was length of the hypoteneuse, but I could not get it to give a reliable answer. Is it possible to do this kind of thing with a neural network?

If so, what would the network look like? just two input cells and one output cell for the answer? What should be in the middle?

$\endgroup$
  • $\begingroup$ Add hidden layers as well and yes it's possible to play with Pythagorean Theorem but if you enter something which is far too from what you trained, that might be erroneous.. it tough to make it generalized for every number out there $\endgroup$ – Aditya Dec 29 '18 at 3:46
  • $\begingroup$ I did add several different configurations of hidden layers with no success, but I was told by someone that neural networks are not able to model higher order relationships of this type. Wondering if that is true or not. $\endgroup$ – user1023110 Dec 29 '18 at 4:04
  • $\begingroup$ Yep it's true but remember that they can map any real valued function to some extent; also you can't use ReLU here probably! An interesting question to explore! $\endgroup$ – Aditya Dec 29 '18 at 4:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.