Advice needed on M -> N regression where N > M

[jamiebullock]

I think this is more of forum question than GH issue, but since the forum is down, maybe someone here can help.

We are trying to do M -> N regression where N > M, e.g. 2 inputs to 8 outputs. Unless the mapping is super-simple, we get some quite strange results with both MLP in regression mode and multiple regression.

My question is: is this even a valid way to use these regression classes, or is there an assumption that number of outputs will always be less than number of inputs.

Thanks.