In regression, you encode a categorical variable with n possible values using n-1 indicator variables. How about for neural networks?
Asked
Active
Viewed 1,215 times
1 Answers
5
You do the same. Although people use $n$ instead of $n-1$ more often, which is probably just a matter of style.

bayerj
- 12,735
- 3
- 35
- 56
-
Why? I'm just starting to learn neural networks, and so far I haven't read about anything that's the equivalent of the intercept in regression. – JenSCDC Jun 24 '14 at 20:40
-
1The intercept is often called bias in neural network texts. – bayerj Jun 25 '14 at 10:03
-
Obviously I still have a good deal to learn about neural networks. – JenSCDC Jun 25 '14 at 13:33
-
I don't think it is only *a matter of style*. See my answer https://stats.stackexchange.com/questions/231285/dropping-one-of-the-columns-when-using-one-hot-encoding/329281#329281 – kjetil b halvorsen Jul 25 '19 at 16:06