Clipped Double Q-Learning

تعلُّم Q المزدوج المقصوص

تمديد لتعلُّم Q المزدوج يستخدم شبكتي هدف ويأخذ الأدنى من تقديراتهما لقيمة الحالة التالية، مما يكبح انحياز المبالغة في التقدير الذي يُصيب تعلُّم Q العادي.

An extension of Double Q-learning that uses two target networks and takes the minimum of their value estimates for the next state, suppressing the overestimation bias that plagues standard Q-learning.

Also translated asQ-Learning المزدوج بالقصّ

First appears in this corpus in: QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation (2018)

Appears in these papers

QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation2018 in the sky ✦
Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor2018 in the sky ✦
TD3: Addressing Function Approximation Error in Actor-Critic Methods2018 in the sky ✦