Unverified50% confidenceFactExact time
AdaBoost中正确分类样本权重乘以e^(-α_t)，错误分类样本权重乘以e^(+α_t)，然后归一化
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAdaBoost中弱学习器权重公式为α_t = (1/2)ln((1-ε_t)/ε_t)，其中ε_t是该轮的加权错误率78% similarUnverifiedAdaBoost的权重更新公式来源于对指数损失函数L=Σ_i e^(-y_i F(x_i))的逐步最小化推导71% similarUnverifiedAdaBoost每一轮训练后提高被错误分类样本的权重，降低被正确分类样本的权重65% similarUnverified贝叶斯决策理论选择使期望损失最小化的行动a*，即a*=argmin_a E[L(θ,a)|x]，期望关于后验分布p(θ|x)计算63% similarUnverified在真实映射为均匀随机的假设下，论文证明了错误率下界为 E ≥ (M/N)δ*(B/M) + (1-M/N)(1-1/K)62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/977408API
curl https://kongchang.com/api/v1/knowledge/claims/977408MCP
get_claim(id=977408)