General Munchausen Reinforcement Learning with Tsallis Kullback-Leibler Divergence

Authors: Lingwei Zhu, Zheng Chen, Matthew Schlegel, Martha White

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show that this generalized MVI(q) obtains significant improvements over the standard MVI(q = 1) across 35 Atari games. [...] We compare MVI(q = 2) with MVI (namely the standard choice where q = 1), and find that we obtain significant performance improvements in Atari. [...] In this section we investigate the utility of MVI(q) in the Atari 2600 benchmark [Bellemare et al., 2013]. [...] Figure 4: Learning curves of MVI(q) and M-VI on the selected Atari games, averaged over 3 independent runs, with ribbon denoting the standard error.
Researcher Affiliation Academia Lingwei Zhu University of Alberta lingwei4@ualberta.ca Zheng Chen Osaka University chenz@sanken.osaka-u.ac.jp Matthew Schlegel University of Alberta mkschleg@ualberta.ca Martha White University of Alberta CIFAR Canada AI Chair, Amii whitem@ualberta.ca
Pseudocode Yes Algorithm 1: MVI(q)
Open Source Code No The paper does not provide an explicit statement about releasing its source code or a direct link to a code repository for the methodology described.
Open Datasets Yes In this section we investigate the utility of MVI(q) in the Atari 2600 benchmark [Bellemare et al., 2013].
Dataset Splits No The paper mentions evaluating on Atari games and performing grid searches for hyperparameters but does not provide specific training/validation/test dataset splits (e.g., percentages, sample counts, or references to predefined splits) needed for data partitioning.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for conducting the experiments.
Software Dependencies No The paper mentions using the 'optimized Stable-Baselines3 architecture [Raffin et al., 2021]' but does not provide specific version numbers for Stable-Baselines3 or any other software dependencies.
Experiment Setup Yes We perform grid searches for the algorithmic hyperparameters on two environments Asterix and Seaquest: the latter environment is regarded as a hard exploration environment. MVI(q) α : {0.01, 0.1, 0.5, 0.9, 0.99}; τ : {0.01, 0.1, 1.0, 10, 100}. Tsallis-VI τ : {0.01, 0.1, 1.0, 10, 100}. [...] Table 1: Parameters used for Gym. [...] Table 2: Parameters used for Atari games.