Unverified50% confidenceFactExact time
1999年Ng、Harada与Russell在ICML证明，只有基于势函数的奖励塑形（F(s,s')=γΦ(s')-Φ(s)）才能保证不改变最优策略
1
Sources
50%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
温室控制器大比拼：强化学习为何败给规则基线？
Related Claims
Unverified基于势函数的奖励塑形（Ng 等人，1999）经理论证明不会改变最优策略，其形式为 F = γΦ(s') - Φ(s)79% similarUnverifiedReward shaping techniques were theoretically analyzed by Andrew Ng and others in a 1999 paper, which proved how reward shaping can be safely performed without altering the optimal policy.62% similarUnverified在纽结理论中，AI发现了双曲体积与代数不变量之间此前未知的函数关系61% similarUnverifiedHardy、Benedicks、Amrein-Berthier等人证明了函数及其傅里叶变换的支撑集不能同时具有有限测度（除非函数为零）60% similarUnverifiedThe CAP theorem was proposed by Eric Brewer in 2000 and formally proved by Gilbert and Lynch in 2002, stating that a distributed system cannot simultaneously guarantee consistency, availability, and partition tolerance.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/563206API
curl https://kongchang.com/api/v1/knowledge/claims/563206MCP
get_claim(id=563206)