Randomised Procedures for Initialising and Switching Actions in Policy Iteration

Authors: Shivaram Kalyanakrishnan, Neeldhara Misra, Aditya Gopalan

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical With the objective of furnishing improved upper bounds for PI, we introduce two randomised procedures in this paper. Our first contribution is a routine to find a good initial policy for PI. ... Our second contribution is a randomised action-switching rule for PI, which admits a bound of p2 lnpk 1qqn on the expected number of iterations. To the best of our knowledge, this is the tightest complexity bound known for PI when k ě 3.
Researcher Affiliation Academia Shivaram Kalyanakrishnan Indian Institute of Technology Bombay Mumbai 400076 India shivaram@cse.iitb.ac.in Neeldhara Misra Indian Institute of Technology Gandhinagar Gandhinagar 382355 India neeldhara.misra@gmail.com Aditya Gopalan Indian Institute of Science Bengaluru 560012 India aditya@ece.iisc.ernet.in
Pseudocode Yes Procedure Guess-and-Max(t) and Algorithm RSPI are provided in the paper.
Open Source Code No The paper does not contain any statement about making source code publicly available, nor does it provide links to any code repositories.
Open Datasets No As a theoretical paper, it does not describe using datasets for training or provide access information for any datasets.
Dataset Splits No As a theoretical paper, it does not involve data splits for validation.
Hardware Specification No As a theoretical paper, it does not involve empirical experiments and therefore does not mention any hardware specifications.
Software Dependencies No As a theoretical paper, it does not describe specific software dependencies with version numbers for experimental reproducibility.
Experiment Setup No As a theoretical paper, it does not describe an experimental setup with hyperparameters or system-level training settings.