Unverified50% confidenceFactExact time
在n维空间中，一个临界点成为局部极小值的概率约为(1/2)^n，因此现代大型网络的几乎所有临界点都是鞍点而非局部极小值
1
Sources
50%
Confidence
Long-term
Relevance
8/22/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在n维空间中一个临界点成为局部极小值的概率约为(1/2)^n，因为要求Hessian矩阵的所有n个特征值都为正77% similarUnverifiedJohnson-Lindenstrauss引理指出在n维空间中可以找到指数级于n的近似正交向量集合61% similarUnverified连续空间中难抵极点的精确解涉及Voronoi图的对偶结构——中轴变换（Medial Axis Transform），难抵极点是中轴上距离值最大的点60% similarUnverified1x1卷积在数学上等价于对每个空间位置独立执行一次全连接变换，不涉及空间邻域信息的聚合59% similarUnverifiedThe Johnson-Lindenstrauss lemma guarantees that for any ε∈(0,1), as long as the target dimension k≥O(log(n)/ε²), there exists a mapping from high dimensions to k dimensions that preserves all pairwise distances within a factor of (1±ε).59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/787195API
curl https://kongchang.com/api/v1/knowledge/claims/787195MCP
get_claim(id=787195)