# M -> N regression where N > M

I am seeking to perform a realtime mapping of M input features to N output parameters where N > M, e.g 2 inputs to 10 outputs. In my use-case I would define regions within my input space and associate parameters in my output space in the training phase, then have the ML infer transitions between these regions in the mapping phase. I guess the opposite of dimensionality reduction!

What would be some good machine learning algorithms for achieving this?