Updated objectives were discussed. Adna will work on ATARI game learning. The goal is to let novices observe RL agents learning performance and to adapt the policy at any point in time through human demonstrations.
We discussed also an schematic overview of the envisioned learning framework.