PID Accelerated Value Iteration Algorithm

Authors: Amir-Massoud Farahmand, Mohammad Ghavamzadeh

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct two sets of experiments. In the first set, we observe the effect of choosing controller gains on the error. In the second set, we study the behaviour of the gain adaptation procedure.
Researcher Affiliation Collaboration 1Vector Institute, Toronto, Canada 2Department of Computer Science, University of Toronto, Canada 3Google Research, Mountain View, California, USA.
Pseudocode Yes Algorithm 1 PID-Accelerated Value Iteration
Open Source Code No The paper does not contain any explicit statement about releasing the source code for the methodology described, nor does it provide a link to a code repository.
Open Datasets Yes We use a chain walk problem with 50 states as a testbed, similar to Lagoudakis & Parr (2003). ... We also use Garnet problems, which are randomly generated MDPs (Bhatnagar et al., 2009).
Dataset Splits No The paper does not provide specific details about training, validation, or test dataset splits. For the MDP problems discussed, the algorithms are typically iterated until convergence rather than using discrete data splits like in supervised learning.
Hardware Specification No The paper describes the algorithms and experiments but does not provide any specific details about the hardware used to run these experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers required to replicate the experiments.
Experiment Setup Yes We set γ = 0.99 in these experiments.