2021
Time-variant variational transfer for value functions
Giuseppe Canonaco, Andrea Soprani, Matteo Giuliani, Andrea Castelletti, Manuel Roveri, Marcello Restelli
UAI 2021poster
In most of the transfer learning approaches to reinforcement learning (RL) the distribution over the tasks is assumed to be stationary. Therefore, the target and source tasks are i.i.d. samples of the same distribution. Unfortunately, this assumption rarely holds in real-world conditions, e.g., due…