Scholarlink

Deep Reinforcement Learning (DRL) has attracted interest for managing intricate decision-making assignments. This article presents a holistic model, Combined DQN with Dueling and Noisy Networks, that combines various DRL innovations. The experimental assessment on CartPole-v1 demonstrates superior performance in scoring and convergence speed when compared to conventional DQN, showcasing the benefits of integrating methods to improve DRL agents. This holistic method signifies a major step forward in reinforcement learning, offering stronger agents for challenging tasks in different fields. The Combined DQN framework takes advantage of Dueling and Noisy Networks to tackle the exploration-exploitation balance and enhance sample efficiency, showcasing its promise for practical use.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Combined DQN with Dueling and Noisy Networks: A Unified Framework for Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • K. LakshmiNadh,
  • S. Siva Nageswara Rao,
  • G. Parimala

摘要

Deep Reinforcement Learning (DRL) has attracted interest for managing intricate decision-making assignments. This article presents a holistic model, Combined DQN with Dueling and Noisy Networks, that combines various DRL innovations. The experimental assessment on CartPole-v1 demonstrates superior performance in scoring and convergence speed when compared to conventional DQN, showcasing the benefits of integrating methods to improve DRL agents. This holistic method signifies a major step forward in reinforcement learning, offering stronger agents for challenging tasks in different fields. The Combined DQN framework takes advantage of Dueling and Noisy Networks to tackle the exploration-exploitation balance and enhance sample efficiency, showcasing its promise for practical use.