Exploiting the short-sighted gradients should be balanced with adequate explorations. Explorations thus should be designed irrelevant to policy gradients in order to guide the policy to unseen states.

Model-free reinforcement learning (RL) algorithms, such as Q-learning, directly parameterize and update value functions or policies without explicitly modeling the environment. They are typically simpler, more flexible to use, and thus more prevalent in modern deep RL than model-based approaches. However, empirical work has suggested that model-free algorithms may require more samples to learn [7, 22]. The theoretical question of "whether model-free algorithms can be made sample efficient " is one of the most fundamental questions in RL, and remains unsolved even in the basic scenario with finitely many states and actions.

Neural Information Processing Systems http://nips.cc/

Neural Information Processing Systems http://nips.cc/

It can be formulated as a reinforcement learning (RL) problem with a known state transition model.

Potential-based reward shaping incorporates prior domain knowledge in the form of additional rewards provided during training to speed up convergence of reinforcement learning algorithms, without changing the optimal policies (Ng et al. [1999]).

It is also prohibitively expensive to learn a new policy from scratch for each robot hardware due to the high sample complexity of modern state-of-the-art algorithms.

Neural Information Processing Systems http://nips.cc/

Neural Information Processing Systems http://nips.cc/

Retouching raw low-quality photos into high-quality ones will greatly increase the aesthetic experience of our vision.