From the source
LG AI Research introduces LobsDICE, a new offline learning from observation algorithm accepted at NeurIPS 2022, which uses stationary distribution correction estimation to mimic expert policy from state-only demonstrations without environment interaction.
