Distributional Hamilton-Jacobi-Bellman Equations for Continuous-Time Reinforcement Learning

Authors: Harley E Wiltzer, David Meger, Marc G. Bellemare

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness of such an algorithm in a synthetic control problem. We simulate the performance of the FD-WGF Q-learning algorithm on a simple task based on a continuous MDP suggested by Munos (2004) as an example of an MDP whose value function does not satisfy the HJB equation in the usual sense.
Researcher Affiliation Collaboration 1Mc Gill University, Montreal, Canada 2Mila Quebec AI Institute 3Google Brain, Montreal, Canada 4CIFAR Fellow.
Pseudocode Yes Algorithm 1 Continuous-time distributional RL update
Open Source Code No The paper does not contain an explicit statement about the release of source code for the described methodology, nor does it provide a link to a repository.
Open Datasets Yes We simulate the performance of the FD-WGF Q-learning algorithm on a simple task based on a continuous MDP suggested by Munos (2004) as an example of an MDP whose value function does not satisfy the HJB equation in the usual sense.
Dataset Splits No The paper does not specify exact percentages, sample counts, or refer to standard predefined splits for training, validation, and test datasets.
Hardware Specification No The paper does not provide specific details such as CPU/GPU models, memory, or cloud computing resources used for the experiments.
Software Dependencies No The paper does not provide specific software dependencies or library versions used for the implementation.
Experiment Setup Yes The discount factor is γ = 0.3, and observations occur at a frequency ω = 1k Hz. ... Algorithm 1 Continuous-time distributional RL update ... Require: WGF time parameter τ Require: Learning rate α