Paulina Tomaszewska
supervisor: Jacek Mańdziuk
Artificial neural networks are inspired by the activity of the human brain. People have the ability to reason about the reality that they observe based on previous knowledge. If they already know how a horse looks like, they do not need a lot of learning to properly distinguish zebra. It is enough to get the information that zebra is like a horse with stripes and then the task is easy. The knowledge transfer in neural networks is performed from different angles. It can be implicit or explicit. The key techniques are: transfer learning, continual learning, knowledge distillation, meta learning and multi-task learning.