Pre-training: A technique used in transfer learning that is based on the idea of training a model on a large amount of data before fine-tuning it to a new task.