Although transfer learning is considered to be a milestone in deep
reinforcement learning, the mechanisms behind it are still poorly understood.
In particular, predicting if knowledge can be transferred between two given
tasks is still an unresolved problem. In this work, we explore th