Regularization and Inference for Practical Reinforcement Learning (RIPL)

Objectives

The RIPL project aims to advance the current state of reinforcement learning by achieving the following objectives:

• (O1) The development of hybrid algorithms that integrate deep learning and symbolic planning in novel ways
• (O2) The advancement of our theoretical understanding of entropic regularization and probabilistic inference in reinforcement learning systems
• (O3) The application of these algorithms and theories in a practical scenario: online social media.

Funding

• Reference: CNS2022-136178
• Start date: 01/09/2023
• End date: 31/08/2025
• Funding organisation: Agencia Estatal de Investigación (AEI)
• Call Details: Programa Plan Estatal de Investigación Científica, Técnica y de Innovación 2021-2023
• Total granted: 168.916,00 €