Unverified50% confidenceFactExact time
Double DQN 用于解决过估计，Dueling DQN 分离状态价值和优势函数，Prioritized Experience Replay 优先采样TD误差大的经验
1
Sources
50%
Confidence
Long-term
Relevance
8/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDQN是off-policy算法，可以重复利用过去收集的经验进行学习66% similarUnverified经验回放（Experience Replay）由DeepMind在2013年DQN论文中引入，将状态-动作-奖励-下一状态四元组存入回放缓冲区，训练时随机采样以打破时序相关性65% similarUnverified在二分类问题中召回率（Recall/Sensitivity）计算公式为TP/(TP+FN)，衡量模型识别所有真实阳性样本的能力63% similarUnverified重复运行以计算通过率比依赖单次结果更能反映模型的真实表现63% similarUnverifiedThe combination of Multi-Token Prediction and speculative decoding can achieve a 2–4x inference speed improvement without compromising generation quality.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/805336API
curl https://kongchang.com/api/v1/knowledge/claims/805336MCP
get_claim(id=805336)