New Ensemble Elastic DQN algorithm reduces overestimation bias in reinforcement learning

Researchers have introduced Ensemble Elastic DQN (EEDQN), a novel reinforcement learning algorithm designed to mitigate overestimation bias in deep Q-networks. EEDQN combines adaptive elastic multi-step returns with ensemble-based target aggregation, using a Q-value difference rule for simpler application in discrete control settings. Evaluations on five MinAtar environments show EEDQN achieving the highest final return in four out of five, outperforming several established DQN variants. AI

IMPACT Introduces a novel approach to improve the stability and performance of value-based reinforcement learning algorithms.

RANK_REASON This is a research paper detailing a new algorithm for reinforcement learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Adrian Łydka
• arXiv
• Averaged DQN
• EEDQN
• Elastic Step DQN
• Ensemble Elastic DQN
• MaxMin DQN

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →