Bayesian Exploration Networks

Authors: Mattie Fellows, Brandon Gary Kaplowitz, Christian Schroeder De Witt, Shimon Whiteson

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results demonstrate that BEN can learn true Bayes-optimal policies in tasks where existing model-free approaches fail.
Researcher Affiliation Academia 1Department of Engineering Science, University of Oxford, Oxford, United Kingdom 2Department of Economics, New York University, New York, United States of America 3Department of Computer Science, University of Oxford, Oxford, United Kingdom.
Pseudocode Yes Algorithm 1 APPROXBRL(PΦ, M(ϕ))
Open Source Code No The paper does not contain an explicit statement about open-sourcing the code for the described methodology or a link to a code repository.
Open Datasets No The paper introduces a novel search and rescue gridworld MDP and evaluates on this custom environment and the Tiger Problem. It does not provide access information (link, DOI, repository, or citation) for a publicly available or open dataset.
Dataset Splits No The paper describes experimental settings like 'episodic' and 'zero-shot' but does not specify explicit training, validation, and test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper states 'The experiments were made possible by a generous equipment grant from NVIDIA' but does not provide specific hardware models (e.g., GPU/CPU models, memory details) used for running experiments.
Software Dependencies No The paper mentions using ADAM for stochastic gradient descent and refers to neural network components (e.g., Re LU activations, gated recurrent unit) but does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes We vary the number of steps for the MSBBE minimisation with a learning rate of 0.02 using ADAM for the stochastic gradient descent. For the for Q-function approximator, we use a fully connected linear layer with Re LU activations, a gated recurrent unit and a final fully connected linear layer with Re LU activations. All hidden dimensions are 32. The dimension of ˆh0 is 2.