Transfer Learning is an AI technique in which a trained model is utilized to address a unique however related issue. Instead of creating the AI machine from scratch the model “transfers” information from a model that has already mastered useful patterns.

Example: Imagine an AI model that is already able to recognize dogs and cats. If you’d like to develop the AI in the recognition of both tigers and lions instead of starting from scratch it’s possible to take the existing model and tweak it using a smaller numbers of data. This can save time, computational power, as well as data.