Transfer Learning with River

[Nirmal-Prabhu]

Hi again to the river team, finding this library very interesting. I have 2 pressing questions which I am trying to get an answer for,

a)I was doing some research on complementing incremental learning with deep learning. As of right now, I am loading the trained weights from a deep learning classifier into a NumPy array or a CSV file and using stream iter to feed this to my streaming pipeline. I understand this acts as a sort of stream learning over trained weights of a deep learning model. Is there a simpler or more iterative way to feed the outputs as the input to my streaming pipeline?

b)Are there any methods through which I can prove that stream learning is taking place at a much faster more efficient rate than deep learning training? Basically, I want to show/quantify its benefits like less computing power and fast training.