Reinforcement Learning
GALOIS: Boosting Deep Reinforcement Learning via Generalizable Logic Synthesis
Figure 2: The illustration of knowledge reused from DoorKey to BoxKey. BoxKey As shown in Figure 1b, different from DoorKey, it has to open the box to get the key. Thus the learned program is color-agnostic (i.e., the agent's policy would remain robust no matter The valuation vector representations are fed to all the methods as input. The reward from the MiniGrid environment is sparse (i.e., only a positive reward will be given after We use a batch size of 256. The code is available at: https://github.com/caoysh/GALOIS
Dynamic allocation of limited memory resources in reinforcement learning
However, the two threads have been largely separate.
PerSim: Data-efficient Offline Reinforcement Learning with Heterogeneous Agents via Personalized Simulators
We perform extensive experiments across several benchmark environments and RL methods.
Reinforcement Learning with Feedback Graphs
We study RL in the tabular MDP setting where the agent receives additional observations per step in the form of transitions samples.
Finite-Time Analysis of Adaptive Temporal Difference Learning with Deep Neural Networks
However, from the theoretical perspective, establishing theoretical convergence guarantees for training DNNs is much more complicated than that for the linear approximation algorithms, which is still widely open.