Hi folks, hope you’re all well.
I was wondering if anyone had any research papers or useful resources regarding design practices for custom head architecture in transfer learning, whether this is task specific (CNN Vision, Recurrent NN’s for NLP etc) or architecture-specific (e.g. other pre-trained NNs like Inception, MobileNet)?
Or even papers which were used as inspiration for the custom head design in fast.ai?
Thanks for any guidance.