As part of a multi-task prediction problem I'm building a neural network for a regression problem where the target/response value lies on the unit interval, i.e. is a real value between 0 and 1. Right now, I'm using a sigmoid output layer (logistic function) to bound the output to the unit interval, but I'm not sure this is actually the best approach. Had I been working in a GLM framework rather than a neural network framework I would probably approach the same problem as a Beta regression, rather than a Logit or Probit. However, a neural network is of course not a linear model, and in some sense "less parametric" than a GLM, so perhaps any activation which bounds the output layer to the unit interval is fine?
Does anyone have any pointers on this? Can I do anything with my neural network to better suit unit interval regression other than have a sigmoid output layer?
