Transfer learning?

Are there examples for transfer learning? I was trying to do that, but realized that it might be impossible with the current Keras nodes for example, even for the same network format. (For example having the trained network and continue the learning.) Did I miss something? Are there (public) examples for transfer learning?
Thanks, gabor

Hi Gabor,

An example for fine tuning can be found here. Please note that this workflow is part of a workflow series, so you’ll have to preprocess the data that is used there using this workflow.

Continuing the learning process (i.e., using the saved optimizer state of a pretrained model) isn’t possible at the moment using the Keras Learner node, unfortunately. Please also refere to this forum thread. You can, however, resort to the deep learning Python scripting nodes. They should preserve the training state.