Experience Replay

إعادة تشغيل التجارب

تقنية تخزّن تجارب الوكيل السابقة (حالة، فعل، مكافأة، حالة تالية) في ذاكرة مؤقتة، ثم تسحب دفعات عشوائية منها للتدريب. تكسر الارتباط الزمني بين العيّنات وتعيد استخدام كل تجربة عدة مرات.

A technique that stores past agent experiences (state, action, reward, next state) in a buffer, then samples random batches for training. Breaks temporal correlation and reuses each experience multiple times.

Also translated asذاكرة إعادة التشغيل، مخزن التجارب المؤقت

First appears in this corpus in: Q-Learning (1992)

Appears in these papers

Asynchronous Methods for Deep Reinforcement Learning2016 in the sky ✦
Asynchronous Methods for Deep Reinforcement Learning2016 in the sky ✦
Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks2017 in the sky ✦
Continuous Control with Deep Reinforcement Learning2015 in the sky ✦
Continuous Control with Deep Reinforcement Learning2015 in the sky ✦
Decision Transformer: Reinforcement Learning via Sequence Modeling2021 in the sky ✦
Deep Reinforcement Learning with Double Q-Learning2016 in the sky ✦
Deep Reinforcement Learning with Double Q-Learning2016 in the sky ✦
Human-Level Control Through Deep Reinforcement Learning2015 in the sky ✦
Human-Level Control Through Deep Reinforcement Learning2015 in the sky ✦
Dueling Network Architectures for Deep Reinforcement Learning2016 in the sky ✦
Dueling Network Architectures for Deep Reinforcement Learning2016 in the sky ✦
First Return, Then Explore2021 in the sky ✦
Hindsight Experience Replay2017 in the sky ✦
IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures2018 in the sky ✦
IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures2018 in the sky ✦
MuZero: Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model2020 in the sky ✦
MuZero: Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model2020 in the sky ✦
Prioritized Experience Replay2015 in the sky ✦
Prioritized Experience Replay2015 in the sky ✦
Q-Learning1992 in the sky ✦
QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation2018 in the sky ✦
Rainbow: Combining Improvements in Deep Reinforcement Learning2018 in the sky ✦
Rainbow: Combining Improvements in Deep Reinforcement Learning2018 in the sky ✦
TD3: Addressing Function Approximation Error in Actor-Critic Methods2018 in the sky ✦