Unverified50% confidenceFactExact time
量化测量域偏移程度的常用方法包括最大均值差异（MMD）和Fréchet Inception Distance（FID）
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
Related Claims
Unverified可借助Fréchet Inception Distance（FID）或Maximum Mean Discrepancy（MMD）等分布距离度量来量化增强后训练分布与真实测试分布之间的偏移80% similarUnverifiedMMD（Maximum Mean Discrepancy）是基于再生核希尔伯特空间的非参数分布距离度量，相比FID对小样本场景更友好、计算开销更低73% similarUnverifiedFID (Fréchet Inception Distance) evaluates generation quality by comparing the statistical distribution differences between real and generated data in the feature space of a deep neural network, with lower values indicating closer resemblance to real data.68% similarUnverifiedFID（Fréchet Inception Distance）由Heusel等人于2017年提出，用预训练Inception网络提取特征并将分布建模为多元高斯，计算Fréchet距离，FID值越低表明分布越接近67% similarUnverifiedProduct quantization converts distance computation from per-dimension floating-point operations into integer table-lookup accumulation using precomputed codebook distances, thereby improving batch retrieval throughput66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/518269API
curl https://kongchang.com/api/v1/knowledge/claims/518269MCP
get_claim(id=518269)