Go to [[Module 1 - Watson AI Overview]] or the [[Main AI Page]]
Using learned patterns from previously-trained models as the basis for new models, layering more specific models on top of less specific ones to speed-up the training of commercially-focused AI applications.
From IBM:
Transfer Learning is learning how to learn. It's what enables Watson to learn more from less, so it doesn't need to be trained from scratch. It can be fed prior knowledge to speed things up.
IBM hints that this is moving towards a solution to the big dataset problem talked about in [[Constraints on AI and robotic advances]].
IBM claims their 3 layer model protects client data in ways other company’s models don’t.
Rendering context...