Unverified50% confidenceFactExact time
格利文科-坎泰利定理保证随着样本量增大，经验分布函数一致收敛到真实分布函数
1
Sources
50%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
Unverified分布的鲁棒性本质上取决于如何建模其尾部行为，用重尾分布替代高斯分布可提升模型对异常值的稳定性75% similarUnverified权重往往近似对称分布适合对称量化，而激活值经过ReLU等非线性函数后分布偏移更适合非对称量化74% similarUnverified后验分布通过贝叶斯定理获得：p(θ|x)∝p(x|θ)·p(θ)，将先验知识与数据似然结合72% similarUnverified经验分布函数的收敛性是 Bootstrap 等重采样方法有效的理论根基71% similarUnverifiedOU过程的平稳分布为高斯分布，其转移核为高斯分布，使任意时刻的条件分布可以精确计算70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/827781API
curl https://kongchang.com/api/v1/knowledge/claims/827781MCP
get_claim(id=827781)