I'd also note that for the case of typical classification that there's an additi...

TTPrograms on March 22, 2016 | parent | context | favorite | on: Some Starting Points for Deep Learning and RNNs

I'd also note that for the case of typical classification that there's an additional concept of "intrinsic boundary dimensionality" that should be low due to the presence of invariants like scale, rotation and translation which drastically lower the intrinsic boundary dimensionality.