Unverified50% confidenceFactExact time
KFAC was originally proposed by Martens and Grosse in 2015 to accelerate natural gradient optimization, with the core idea of approximating the Fisher information matrix of each layer as a Kronecker product of two smaller matrices.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedK-FAC utilizes the Kronecker decomposition structure of the Fisher information matrix to compress the originally O(n²) curvature matrix into the product of two smaller matrices.74% similarUnverifiedMuon优化器由Kosson等研究者提出，核心改进在于对梯度更新矩阵进行正交化处理67% similarUnverified双线性池化方法约在2015年前后出现，是FGVR的重要里程碑，通过两个特征网络输出的外积捕捉二阶统计关系61% similarUnverifiedGPTQ is a layer-wise quantization method based on approximate second-order information, proposed by Frantar et al. in 2022, which minimizes quantization reconstruction error through the approximate inverse of the Hessian matrix.61% similarUnverifiedCovariance matrices are inherently positive semi-definite, and this property is the theoretical foundation of Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA)61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001291API
curl https://kongchang.com/api/v1/knowledge/claims/1001291MCP
get_claim(id=1001291)