Sequential Transfer in Reinforcement Learning with a Generative Model

Authors: Andrea Tirinzoni, Riccardo Poiani, Marcello Restelli

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we empirically verify our theoretical findings in simple simulated domains.
Researcher Affiliation Academia 1Politecnico di Milano, Milan, Italy. Correspondence to: Andrea Tirinzoni <andrea.tirinzoni@polimi.it>.
Pseudocode Yes Algorithm 1 Policy Transfer from Uncertain Models (PTUM)
Open Source Code No The paper does not provide any statement or link indicating that source code for the described methodology is publicly available.
Open Datasets Yes the 4-rooms domain of Sutton et al. (1999) with only two rooms)
Dataset Splits No The paper describes experimental setups with different tasks and domains (e.g., grid-world, object-world) but does not refer to traditional dataset splits like 'training', 'validation', or 'test' sets in the context of supervised learning.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as CPU/GPU models or memory specifications.
Software Dependencies No The paper does not provide specific version numbers for any software dependencies or libraries used in the experiments.
Experiment Setup Yes We adopt a standard 12 12 grid-world divided into two parts by a vertical wall (i.e., the 4-rooms domain of Sutton et al. (1999) with only two rooms). The agent starts in the left room and must reach a goal state in the right one, with the two rooms connected by a door. We consider 12 different tasks, whose models are known to the agent, with different goal and door locations.