Unverified50% confidenceFactExact time
单变量线性回归模型可表示为 y = w * x + b，其中w是权重（斜率），b是偏置（截距）
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified多变量线性回归可用矩阵形式表示为 y = X · W + b80% similarUnverified低秩参数化将完整协方差矩阵分解为Σ = WW^T + D的形式，使参数量从O(d²)降至O(dr)67% similarUnverifiedL2正则化（Ridge回归）从贝叶斯视角等价于假设模型参数服从均值为零的高斯先验分布63% similarUnverifiedLoRA的原理是将微调中权重变化量ΔW分解为两个低秩矩阵的乘积（ΔW = A×B）63% similarUnverified如果所有层都仅做线性变换，无论网络多少层其整体效果等价于单层线性模型63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/901603API
curl https://kongchang.com/api/v1/knowledge/claims/901603MCP
get_claim(id=901603)