Unverified50% confidenceFactExact time
The Gauss-Markov theorem states that when the error terms satisfy the three conditions of zero mean, homoscedasticity, and no autocorrelation, the OLS estimator is the Best Linear Unbiased Estimator (BLUE).
1
Sources
50%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
Unverified线性回归的最小二乘损失与逻辑回归的交叉熵损失分别对应高斯噪声假设与伯努利分布假设下的最大似然估计68% similarUnverified贝叶斯最优分类器在0-1损失下等价于选择后验概率最大的类别，回归问题中条件期望在平方损失下是最优预测62% similarUnverified当z>4时，在零假设下观察到如此极端值的概率约为p≈3.2×10⁻⁵，即每十万次检测中约3次误报62% similarUnverifiedAssuming the data follows a Gaussian distribution, maximizing the likelihood is equivalent to minimizing the mean squared error.61% similarUnverified困惑度是测试集上每个token平均负对数似然的指数，值越低代表模型对语料的预测越准确60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/829233API
curl https://kongchang.com/api/v1/knowledge/claims/829233MCP
get_claim(id=829233)