Unverified50% confidenceTradeoffExact time
Ledoit-Wolf shrinkage is an application of the bias-variance trade-off, introducing moderate bias to significantly reduce estimation variance and thereby minimize overall estimation error
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified正则化方法（如L1/L2正则化、Dropout、早停法）本质上是通过引入少量偏差来大幅降低方差，从而提升泛化能力66% similarUnverifiedTotal prediction error can be decomposed into bias² + variance + irreducible error.64% similarUnverified对MoE模型的量化需要格外谨慎，因稀疏激活使专家参数数值分布集中，低比特量化易造成路由偏差导致能力退化64% similarVerifiedModern quantization schemes such as GPTQ and AWQ introduce layer-by-layer calibration and error compensation mechanisms to keep perplexity loss within 1–2%64% similarVerifiedGradient descent iteratively adjusts weights in the opposite direction of the loss function's gradient, causing the prediction error to converge progressively.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/891773API
curl https://kongchang.com/api/v1/knowledge/claims/891773MCP
get_claim(id=891773)