High Entropy and Lower Dataset count Data regression and classification using DNNs

How to deal with Datasets that have more feature values than Label count and also those data that have more classifications using DNNs. The example applications being NIR Spectroscopy and Visual Perception analysis using relatively large symbol sets represented using few pixels.