Unverified50% confidenceSolutionExact time
聚类前可做L2归一化或改用以余弦距离为度量的Spherical K-Means变体来解决模长主导问题
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified两个不可约表示的张量积l₁⊗l₂可分解为|l₁-l₂|到l₁+l₂之间所有整数阶不可约表示的直和，CG系数给出具体的基变换矩阵69% similarUnverified当特征向量被L2归一化后，余弦相似度等价于点积运算68% similarUnverified对已做L2归一化的嵌入，余弦距离与欧氏距离给出的排序结果完全一致64% similarVerified从贝叶斯视角理解，L1正则化对应参数的拉普拉斯先验，L2正则化对应高斯先验63% similarUnverifiedL2算法相对于基向量系数位长β达到O(n⁴·β·(n+β))的位复杂度，在β=O(n)时相当于O(n⁶)，而经典LLL约为O(n⁹)60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/901371API
curl https://kongchang.com/api/v1/knowledge/claims/901371MCP
get_claim(id=901371)