Should I bin my continuous input variables for neural networks?

Asked Sep 02 '13 at 14:08

Active Sep 02 '13 at 14:09

Viewed 131 times

And what is the optimal method to do this? Is there a package for this in R?

edited Sep 02 '13 at 14:09

Nick Cox

asked Sep 02 '13 at 14:08

user2298382

2

See [here](http://stats.stackexchange.com/questions/68834/what-is-the-benefit-of-breaking-up-a-continuous-predictor-variable/68839#68839). And I don't think using a neural net makes any difference. – Scortchi - Reinstate Monica Sep 02 '13 at 14:34
2

Using neural nets does make a difference. See http://cs231n.github.io/neural-networks-2/#losses under subsection 'word of caution'. One of the best researchers within neural networks stating: "It is important to note that the L2 loss is much harder to optimize than a more stable loss such as Softmax. (...) When faced with a regression task, first consider if it is absolutely necessary. Instead, have a strong preference to discretizing your outputs to bins and perform classification over them whenever possible." – pir Jun 30 '15 at 22:09

0 Answers0