approved
CPG: Constrained Policy Gradient

CPG is a reinforcement learning algorithm thought for solving continuous control problems with user defined or structural constraints. CPG has two versions, C-PGPE and C-PGAE, based on the kind of exploration the user want to perform (the former is w.r.t. the parameters, the latter w.r.t. the actions).

Tags
Data and Resources
To access the resources you must log in
Additional Info
Field Value
Accessibility Both
AccessibilityMode Download
Associate Project FAIR
Associate Project FAIR
Basic rights Other rights
CreationDate 2025-04-08
Creator Montenegro, Alessandro, alessandro.montenegro@polimi.it, orcid.org/0009-0000-2034-7106
Field/Scope of use Research only
Group Others
Owner Montenegro, Alessandro, alessandro.montenegro@polimi.it, orcid.org/0009-0000-2034-7106
Programming Language Python
RelatedPaper Alessandro Montenegro, Marco Mussi, Matteo Papini, Alberto Maria Metelli. Last-Iterate Global Con- vergence of Policy Gradients for Constrained Reinforcement Learning. 2024. NeurIPS, Neural Information Processing System
SoBigData Node SoBigData EU
Sublicense rights No
Territory of use World Wide
Thematic Cluster Other
system:type Method
Management Info
Field Value
Author Montenegro Alessandro
Maintainer Montenegro Alessandro
Version 1
Last Updated 22 June 2025, 01:03 (CEST)
Created 22 June 2025, 01:03 (CEST)