Variational Bayesian Optimistic Sampling

Authors: Brendan O'Donoghue, Tor Lattimore

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

Reproducibility Variable Result LLM Response
Research Type Experimental In Figure 2 we show how five agents perform on a 50 50 randomly generated game in self-play and against a best-response opponent.
Researcher Affiliation Industry Brendan O Donoghue Deep Mind bodonoghue@deepmind.com Tor Lattimore Deep Mind lattimore@deepmind.com
Pseudocode Yes Algorithm 1 TS for bandits
Open Source Code No 3. If you ran experiments... (a) Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No] We included a description of the data generation process for the simulations we ran.
Open Datasets No The entries of R were sampled from prior N(0, 1) and the noise term ηt at each time-period was also sampled from N(0, 1).
Dataset Splits No The paper describes data generation processes (e.g., 'entries of R were sampled from prior N(0, 1)') but does not specify explicit train/validation/test dataset splits for reproducibility.
Hardware Specification No 3. If you ran experiments... (d) Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [Yes] In the appendix. (Specific details are not present in this excerpt.)
Software Dependencies No The paper does not explicitly provide specific software dependencies with version numbers needed to replicate the experiment.
Experiment Setup Yes The entries of R were sampled from prior N(0, 1) and the noise term ηt at each time-period was also sampled from N(0, 1). ... For this experiment we set C = 10... We compare the K-learning and UCB algorithms...over 8 seeds.