… Policy Optimization from Demonstration (POfD) method, which can acquire knowledge from
demonstration … 1) We reformulate the policy optimization objective by adding a demonstration…

… HYbrid Policy Optimization (HYPO), which uses a small number of imperfect demonstrations
… The key idea is to train an offline guider policy using imitation learning in order to instruct an …

… the demonstrations … the policy only specifies a distribution over the action space. We force
the actions of the policy to equal the demonstration actions instead of sampling from the policy …

… In this work we develop a new guarded policy optimization … In contrast to the offline learning
from demonstrations, in this work we focus on the online deployment of teacher policies with …

… Policy Optimization with Smooth Guidance (POSG) that leverages a small set of sparse-reward
demonstrations to … be indirectly estimated using offline demonstrations rather than directly …

… introduce the Iterative Regularized Policy Optimization (IRPO) method … demonstration
boosting to enhance demonstration quality. Specifically, iterative training capitalizes on the policy …

… 2021) analyze the constrained pessimistic policy optimization with general function
approximation and with the partial coverage assumption of the dataset, then they specialize the case …

… constraints which guide the optimization performed by Approximate Policy Iteration. We …
expert demonstrations. In all cases, we compare our algorithm with LeastSquare Policy Iteration (…

… On the contrary, we show that for a Diffusion Policy pre-trained from expert
demonstrations, our methodology for fine-tuning via PG updates yields robust, high-performing …

… Our EPG algorithm is derived and approximated from the optimization of perturbed policies
with policy gradient methods, by adding some heuristic of EAs. As EAs differ in how to …