Document type
OtherVersion
Published versionPublication date
Publication license
Please use this identifier to cite or link to this item: https://hdl.handle.net/2445/182959
Towards sample-efficient policy learning with DAC-ML
Journal Title
Director/Tutor
Journal ISSN
Volume Title
Related resource
Abstract
The sample-inefficiency problem in Artificial Intelligence refers to the inability of current Deep Reinforcement Learning models to optimize action policies within a small number of episodes. Recent studies have tried to overcome this limitation by adding memory systems and architectural biases to improve learning speed, such as in Episodic Reinforcement Learning. However, despite achieving incremental improvements, their performance is still not comparable to how humans learn behavioral policies. In this paper, we capitalize on the design principles of the Distributed Adaptive Control (DAC) theory of mind and brain to build a novel cognitive architecture (DAC-ML) that, by incorporating a hippocampus-inspired sequential memory system, can rapidly converge to effective action policies that maximize reward acquisition in a challenging foraging task.
Subject (English)
Citation
Citation
FREIRE, Ismael T., et al. Towards sample-efficient policy learning with DAC-ML. Procedia Computer Science. 2021. Vol. vol 190, num. 256-262. ISSN Freire IT;Amil AF;Vouloutsi V;Verschure PFMJ. Towards sample-efficient policy learning with DAC-ML. Procedia Computer Science, 2021, 190, 256-262. [consulted: 11 of August of 2026]. Available at: https://hdl.handle.net/2445/182959