Inverse reinforcement learning attempts to reconstruct the reward function in a Markov decision problem, using observations of agent actions. As already observed in Russell [1998] the problem is ill-posed, and the reward function is not identifiable, even under the presence of perfect information about optimal behavior. We provide a resolution to this non-identifiability for problems with entropyregularization.

Thecentral nodeisthe busiest node since it needs to communicate with each worker concurrently.

Weimplementedthree algorithms: Centralized Parallel Optimistic Adam (CP-OAdam), Decentralized Parallel Optimistic Adam (DP-OAdam) and Randomization Decentralized Parallel Adam (Rand-DP-OAdam) inspired by [20,21].

We 39] UnivCon hasserv (real-vf and g, we examine filter. Wechoose Thai Statue, Armadillo, and Dragonfrom Stanford 3DScanning Repository [84,85,86,87] todemonstrateourresults. Figure 1 8 Input Image Mean Filter Median Filter LaMaINSP-Net Target Image